To find out whether that missing information was an accident of digitisation or a habit of the discipline itself, the researchers went back to the original publications. They searched vegetation databases, the Scopus literature database, reference lists and personal inquiries to colleagues, gathering 226 studies and analysing 172 of them. They restricted the scope to Europe, where both traditions have long histories, and kept only studies based on plot observations in arable fields.

For each paper they recorded what was said about study design, plot position and size, survey timing, abundance measurement, and the environmental and management variables collected alongside the plant data. Sorting the studies into disciplines proved to be a challenge in itself, since papers rarely announce which community their authors belong to. The researchers tried two approaches.

The first classified each journal by its stated scope, along a gradient from specialised phytosociology and vegetation science, through botany and general ecology, to agronomy and specialised weed science. The second classified each study by its stated aims: phytosociological description of plant communities, regional inventories of weed flora, or analyses of the factors influencing weed vegetation, the last split into purely environmental drivers and those combining management with environment. Both classifications, the authors note, describe a gradient rather than a strict divide, with considerable overlap between the communities.

The results show a consistent pattern. Of four major methodological variables — plot position in the field, survey season, plot size and abundance measure — the abundance scale was reported in 93 percent of all studies and plot size in 82 percent, but survey season appeared in only 67 percent and the surveyed field part in just 49 percent. When broken down by study aim, the differences sharpened.

Studies analysing the influence of environment and management on weed vegetation reported the highest number of methodological details, while phytosociological studies reported the fewest. Abundance measures were reported equally across all categories, but the other three aspects were most often documented in management-focused work and least often in phytosociology. Phytosociological studies in the review reported plot sizes starting from around 10 square metres, consistent with long-standing recommendations in vegetation science, whereas studies pursuing agronomic questions sometimes worked with plots far smaller than 10 square metres — in extreme cases down to fractions of a square metre, repeated many times within a field to capture the patchiness of sparse weed stands.

The review notes that recommendations for arable weed communities have ranged from under 16 square metres to as much as 1,000 or 2,000 square metres, and that practical plot sizes in weed science have spanned from 0.1 square metres to 20,000 square metres. Species richness, as ecological theory and decades of empirical work confirm, depends directly on the area sampled, so these choices translate directly into the numbers a survey records. The Braun-Blanquet approach, the dominant cover-estimation method in European phytosociology, illustrates the reporting problem neatly.

Some studies in the review cited the classic 1964 handbook and considered their methodology sufficiently described by that reference alone. But the handbook does not prescribe a plot size for arable communities, and among the phytosociological studies that did state a plot size, values ranged from 1 to over 100 square metres. A reader encountering only the words "Braun-Blanquet method" therefore cannot reconstruct what was actually measured — a serious obstacle when the data are later pooled with observations made under entirely different sampling regimes.

The differences extend beyond plot size. Phytosociological surveys tend to be timed later in the growing season, often spanning spring to summer to capture the peak of vegetation development, while agronomically motivated surveys are frequently carried out in spring, and occasionally even in autumn — an unusual season for vegetation science, but a natural one for weed scientists, who often assess weed seedlings in winter crops before herbicide decisions are made. Field position matters too: weed vegetation differs between field centres and edges because tillage, fertiliser and herbicide intensity typically decline towards field boundaries, and machinery turning areas suffer soil compaction.

Studies of management effects therefore often explicitly sample at a fixed distance from the field limit to avoid edge effects, and they report this choice. Phytosociological studies, the review found, reported the surveyed field part least often of all categories. The consequences surface when data are reused.

Modern vegetation science increasingly relies on large compiled databases, and ecologists have developed workable solutions for many compatibility problems, such as converting between different cover scales or between abundance measures. Missing methodological metadata are harder to fix: if a database entry does not say whether the plot sat in a field centre or at its margin, that information usually cannot be recovered from the species list itself. The review also highlights sampling design as an overlooked variable — an inventory aiming at complete regional coverage will sample differently from a study comparing organic and conventional farms, and preferential versus stratified or randomised plot selection will shape results in ways that only become interpretable if the design was documented.

The authors' recommendation is a dedicated reporting standard for weed vegetation surveys, developed through an interdisciplinary process. They urge future surveyors to record and publish five details alongside every plot observation: plot size, plot position relative to the field edge, the date or at least month of recording, the abundance measure used, and the crop grown in the field. They also call on those digitising published data to carry as many of these variables as possible into electronic databases, using Latin or English crop names to ease translation.

Recording such variables in the field, they argue, would cost little time or money while adding lasting value.

Subject of Research: Methodological reporting differences between vegetation science and weed science surveys of arable weed vegetation in Europe

Article Title: Vegetation survey methodology in arable weeds is reported with less detail from vegetation science than weed science

Article References: Bürger, J., & Küzmič, F. (2023). Vegetation survey methodology in arable weeds is reported with less detail from vegetation science than weed science. Vegetation Classification and Survey, 4, 209-218. https://doi.org/10.3897/vcs.105300

Image Credits: AI Generated

DOI: 10.3897/VCS.105300

Keywords: arable weeds, vegetation science, weed science, phytosociology, survey methodology, plot size, Braun-Blanquet, biodiversity databases, Europe, agriculture, data standards, field surveys

This is a methods study based on the authors’ reflective analysis of six large landholder surveys commissioned by the Cooperative Research Centre for High Performance Soils, also known as the Soil CRC. It draws on three decades of social benchmarking in Australia and combines survey instruments and drafts, internal notes, audits, regional reports and feedback from partner organisations. The surveyed areas were North Central and Wimmera in Victoria, the Eyre Peninsula of South Australia, Central West New South Wales, mainland Tasmania and the Wheatbelt of Western Australia.

The authors describe the approach as a way to produce comparable evidence about landholders’ values, beliefs, norms and management practices while retaining sensitivity to local conditions. It was not a controlled trial of any farming practice, and the reported relationships do not demonstrate that changing an attitude, value or information source would cause adoption. At the centre of the method is a questionnaire structured around 12 areas of inquiry: occupational identity, enterprise and land-use mix, salient issues, general value orientations, place-based values, knowledge, confidence and capacity to act, trusted information networks, attitudes to risk, trust and climate, on-farm management actions, demographic and farm context, and long-term intentions.

The design integrates several behavioural theories. Ajzen’s Theory of Planned Behaviour links intention to behaviour through attitudes, perceived social norms and perceived behavioural control. Value-Belief-Norm theory connects environmental behaviour to values, beliefs and personal responsibility, while the Assigned Values Model examines the relationship between a landholder and the particular place being managed.

The researchers distinguish broadly held principles, such as creating wealth or living in harmony with nature, from assigned values attached to features such as soil, identity, family, community or the future condition of a property. Rather than using an identical questionnaire everywhere, the method retains stable “sister questions” for longitudinal and cross-regional comparison while adding locally selected practices and issues. Regional partners map priorities in workshops, after which the instrument is tailored, pre-tested with local landholders and revised for clarity and relevance.

Most questions use five-point Likert scales, multiple-choice options or categorical responses, while open questions allow respondents to raise issues not anticipated by researchers. A “Do not know/Not applicable” option helps distinguish uncertainty or inapplicability from disagreement. Open responses in one regional instrument identified mobile connectivity, long-range weather forecasting and soil-management knowledge as practical needs that linked several otherwise separate parts of the survey.

Sampling is designed to support broad inference. Researchers obtained spatially referenced lists of properties larger than 10 hectares, then surveyed all eligible landholders in sparsely populated regions or randomly selected the required number in larger ones. The preferred mail-out sample was 2,000, with approximately 250 usable responses treated as generally sufficient for descriptive summaries, exploratory pairwise analyses and follow-up modelling.

Under common survey assumptions, this size can produce margins of error of about 5% to 7%, although actual precision depends on the sampling design and population characteristics, for which formal probability-based calculations are required. The team also compared average land size among respondents with broader regional landholder data as a practical check for skew toward large or small properties, although this check cannot identify every form of non-response bias. Delivery followed the systematic approach associated with Dillman’s Tailored Design Method.

An advance notice informed recipients that the survey was coming, offered an opt-out and provided a QR code for online completion. Those who did not opt out received a printed questionnaire and cover letter, followed by two mailed reminders, another survey and a final reminder. From the second survey onward, a regional online version linked responses to mailed serial numbers.

Local promotion through radio, farming events, media releases and partner networks was treated as essential. Reasons for non-participation included the belief that the survey did not apply to the respondent, lack of time, poor health or age, and a general dislike of surveys. Response rates ranged from 24% in the Western Australian Wheatbelt to 36% in North Central Victoria.

In the Wheatbelt, a state closure prevented the project lead from promoting the survey in person. A less intensive 2023 Wimmera trial that sent flyers with QR codes to 2,000 landholders in two local government areas produced an extremely low response and disproportionately greater participation by women and non-farmers. The analysis proceeded from description to targeted modelling.

Researchers calculated frequencies, means and medians using the available responses for each question, so denominators varied and “Do not know/Not applicable” answers and missing values were excluded from item-level summaries. For reporting, responses of 1–2 and 4–5 on Likert scales were grouped as negative and positive categories, while the original scale values were retained for pairwise comparisons and modelling. Linear regression was used for continuous outcomes, logistic regression for binary outcomes and appropriate categorical models for outcomes with more than two classes.

In each region, as many as 20–40 prioritised associations from pairwise screening were advanced to regression. A p-value below 0.05 was used as the screening threshold, but significance was evaluated alongside sample availability, missing data, practical relevance, local knowledge and consistency with the framework. Spatially referenced landholder data also made it possible to map patterns against features such as soil type.

Illustrative results from a 2025 report for North Central Victoria showed that relationships differed across practices. Whole-farm planning was most strongly associated with knowledge of allocating land use according to land class, together with knowledge and implementation of soil-health, biodiversity and grazing approaches. A multivariable model linked whole-farm planning with business-planning knowledge, use of time-controlled, holistic or cell grazing, management of a larger area, succession planning and longer decision horizons.

The model reported Cox’s Snell R² of 0.40, Nagelkerke R² of 0.54, 85 degrees of freedom and p = 2.03 × 10⁻⁶. Time-controlled grazing was associated with knowledge and confidence in the practice, reduced synthetic use per hectare, succession planning and an assigned value of contributing to positive agricultural change. That model reported Cox’s Snell R² of 0.43, Nagelkerke R² of 0.58, 97 degrees of freedom and p = 5.54 × 10⁻¹⁰.

These are observational associations, not estimates of intervention effects. Values also organized distinct sets of priorities. “Creating wealth” was associated with 24 salient issues, mostly concerning soil but also water security, social licence and missing services. It was positively linked with knowledge of farming and soil health, trusted advisers and several farm practices, but negatively associated with using environmental organisations for information and with understanding of climate change. “Living in harmony with nature” was linked with 23 issues including soil health, biodiversity, water and climate change, as well as belief in science, data-driven decisions and confidence in regenerative agriculture.

Its model included a desire to leave the property in better condition, concern about changing weather and a normative view concerning First Nations access to cultural sites on private land, with R² of 0.35, 263 degrees of freedom and p = 4.85 × 10⁻²³. Overall, the authors reject a deterministic sequence in which values simply produce beliefs, norms and then behaviour: values can shape which organisations landholders trust, trusted networks can reinforce or weaken openness to new information, and one practice may increase the likelihood of adopting another. The method’s central practical lesson is that representativeness depends on sustained local collaboration as much as on technical sampling.

Partners help define meaningful regional boundaries, identify priorities, supply contact information, promote the survey and connect findings to planning cycles. Repeated application can provide evidence about locally associated factors such as knowledge, capacity constraints, trusted sources and assigned values while retaining a comparable core for wider analysis. The authors caution that the framework is not yet a deterministic predictive tool and that further research is required to determine its full applications.

The datasets are also not openly available because of ethics requirements. Even with those limitations, implementation across six regions shows how a stable theoretical core can be adapted locally without sacrificing longitudinal comparability, offering a practical model for tracking social change in land management from local programs through to national policy.

Subject of Research: A repeated social survey method for examining factors associated with rural landholder decisions and farming-practice adoption in Australia

Article Title: A Social Benchmarking Method for surveying rural landholders

Article References: Luke, H., Allan, C., Alexanderson, M., & McDonald, S. (2026). A Social Benchmarking Method for surveying rural landholders. Frontiers in Sustainable Food Systems, 10, Article 1746415. https://doi.org/10.3389/fsufs.2026.1746415

Image Credits: AI Generated

DOI: 10.3389/fsufs.2026.1746415

Keywords: social benchmarking, rural landholders, agricultural surveys, practice adoption, behavioural research, land management, survey methodology, Australia, soil health, trusted networks, Social, Benchmarking

The problem Katz targets is what he calls the accidental scale. When a survey module asks respondents whether they are nervous or excited about AI, whether AI threatens their job, whether AI companies should be regulated, and whether the federal government is equipped to regulate the technology, analysts and journalists often average these answers into one index of public sentiment. The practice is tempting because the items are topically connected, but topical connection is not the same as measuring the same thing. A person can feel uneasy about AI, fear for their job, still want strict regulation, and doubt that Washington has the technical competence to deliver it. Those are four distinct judgments about four distinct objects.

To test whether the items really belonged together, Katz used principal-component analysis on polychoric correlations, a technique suited to ordinal survey responses, on the complete-case subset of 460 respondents who answered all four attitude and policy questions. The results were unambiguous. The Kaiser-Meyer-Olkin statistic, a standard measure of sampling adequacy for common-factor analysis, came in at 0.48, below the 0.50 threshold that Kaiser himself classified as unacceptable and well below the 0.60 conventionally required for adequate common structure. Two components had eigenvalues above one, and the government-capacity item loaded almost entirely on its own component, near 0.96, with negligible cross-loading on the sentiment component.

Reliability was equally poor. The ordinal alpha for the three sentiment-related items was just 0.45, far below the 0.60 floor typically considered inadequate for scale scoring. Oblique rotations using promax and direct-oblimin methods confirmed that the separation was not an artifact of forcing components to be orthogonal; inter-component correlations stayed below 0.07. A companion sensitivity analysis on the three non-employment items, estimated on a larger complete-case sample, reproduced the same two-component pattern, showing that the negative finding was not driven by who happened to be asked the job question. The weak first component that did emerge linked excitement, lower job concern, and opposition to regulation, but with reliability this low it cannot support meaningful scale scoring.

The government-capacity item deserves particular scrutiny. When respondents were asked whether the U.S. federal government is equipped to regulate AI if it decides to do so, their answers correlated more strongly with presidential approval and with a battery assessing administration performance than with the AI affect item. The correlation with the administration-performance battery was 0.28, and with presidential approval 0.25, compared with much weaker links to AI sentiment. This suggests the question is functioning partly as a measure of confidence in the incumbent government rather than as a pure indicator of attitudes toward artificial intelligence. In a partisan environment, asking whether a named institution can handle a named technology inevitably blends two objects of evaluation.

Having established that no composite index was defensible, Katz turned to the single item with the broadest theoretical reach: the nervous-to-excited question, which was asked of the full sample and captures a global affective appraisal of the kind risk-perception research treats as the broadest summary of orientation toward an unfamiliar hazard. Weighted regression with HC3 heteroskedasticity-robust standard errors, corroborated by multiple imputation across 20 datasets, revealed a clear hierarchy of predictors. Personal AI use was the strongest correlate, followed by prospective economic optimism, age, and education. The ordering held in both the listwise specification, where AI use carried a standardized coefficient of 0.33, and the multiply imputed specification, where it was 0.31.

The age result is especially instructive. At the zero-order level, age showed no significant association with AI excitement. But once AI use and other predictors were controlled, age became positively associated with excitement, a suppression pattern that appeared in both estimation strategies. Younger respondents use AI more frequently, and use is positively associated with excitement. Holding use constant, older respondents actually score somewhat higher on the affect item. The popular narrative that older people are uniformly more fearful of AI does not survive this multivariate scrutiny, and apparent generational differences may partly reflect unequal exposure rather than deep dispositional gaps.

The descriptive picture from the survey is itself revealing. About 70 percent of respondents were nervous rather than excited about recent AI advances, while roughly one in five sat on the excited side. Roughly two thirds reported using AI at least occasionally, but only 18 percent used it every day or most days. Support for closer federal regulation of AI companies was broad at near 78 percent, yet confidence that the federal government is equipped to regulate AI was low at near 19 percent. Among employed respondents, about one third worried that AI could threaten their job. These are not facets of a single mood; they are distinct evaluations that happen to share a topic.

Katz is careful to bound his claims. The analysis covers one survey module in one state, drawn from registered voters who skew older and more racially homogeneous than many national populations, and the cross-sectional design cannot establish whether use causes favorability or favorable respondents simply select into use. The affect outcome is a single bipolar item without a neutral category, which may compress ambivalence. The dimensionality result rests on complete cases, and while the three-item companion analysis addresses the most consequential selection concern, the finding should be read as a bounded negative result about this item set, not as proof that public attitudes toward AI are intrinsically fragmented. Purpose-built instruments, including validated multidimensional scales and reliable brief unidimensional measures, exist in the psychometric literature and continue to be refined.

The practical implications, however, are hard to escape. Survey organizations should test dimensionality before building an AI-attitude index, separate behavioral use from evaluative attitude unless theory explicitly joins them, and distinguish policy preference from institutional trust in governance batteries. Support for regulating AI companies and confidence that a particular government can regulate them are conceptually different judgments, and combining them blurs the line between wanting oversight and trusting the available overseer. As AI moves from specialist concern to routine political battleground, the temptation to report a single public mood will only grow. Katz’s warning is that even a small module of four topically related items can behave as an accidental scale, and that headline numbers built from such scales can be numerically precise while being conceptually incoherent. The accidental scale should be diagnosed and avoided, not treated as an implicit default for applied polling on AI.

Subject of Research: Testing the dimensionality of public attitudes toward artificial intelligence

Article Title: Testing the dimensionality of public attitudes toward artificial intelligence

Article References: Katz, J. E. (2026). Testing the dimensionality of public attitudes toward artificial intelligence. Discover Artificial Intelligence, 6(1), Article 1518. https://doi.org/10.1007/s44163-026-02338-1

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02338-1

Keywords: artificial intelligence, public opinion, survey methodology, psychometrics, AI governance, risk perception, digital divide, attitude measurement, Testing, dimensionality, public, attitudes

The study, led by Suricha Artamaprasangsa with colleagues Sudarat Tuntivivat and Polthep Poonpol, began from a deceptively simple observation. Existing grit measures, most famously the scale developed by psychologist Angela Duckworth, were designed largely around individual, self-directed pursuits such as education, military training, or competitive pursuits. Manufacturing work is different in kind. On an automotive assembly line, effort is synchronized with machines and colleagues, goals are collective rather than personal, and long-term commitment is expressed through attendance, precision, and adherence to disciplined routines rather than through individual ambition. The researchers argued that a generic questionnaire might fail to detect the distinctive behavioral signatures of grit in this environment, and that a context-specific instrument was needed before anyone could meaningfully measure, predict, or cultivate persistence among production workers.

To build that instrument, the team adopted a sequential exploratory mixed methods design, an approach that deliberately puts qualitative discovery first and statistical confirmation second. In the qualitative phase, eight supervisors and senior supervisors with between 20 and 35 years of production experience took part in semi-structured interviews. These were not casual conversations; they were structured probes into what sustained effort and long-term commitment actually look like in observable behavior on the line. The researchers then applied a hybrid deductive-inductive thematic analysis, a technique that both tests existing theoretical categories and allows new themes to emerge from the data itself. Crucially, the analysis retained the two established dimensions of grit, perseverance of effort and consistency of interest, while also surfacing ten contextual behavioral indicators that generic scales had never captured.

Those ten indicators form the conceptual heart of the new scale. Drawing on the interview material, the researchers wrote two questionnaire items for each indicator, producing an initial pool of 20 items. Before any large-scale testing, the pool was subjected to two rounds of scrutiny. First, five independent experts assessed each item for congruence with its intended construct and for clarity of wording, a standard procedure in scale development designed to weed out ambiguous or double-barreled questions. Second, 30 automotive operators completed a pilot version of the scale, providing early evidence about how the items functioned with the very population they were written for. Only after this filtering did the team move into the quantitative phase, in which 204 operational employees completed the full instrument.

The statistical analysis that followed was rigorous and transparent. Five items were removed based on combined evidence from the pilot test and a confirmatory factor analysis, the technique psychologists use to test whether a proposed structure of a questionnaire actually matches the pattern of responses people give. The researchers then estimated the remaining model using diagonally weighted least squares, an estimation method well suited to the ordered categorical response formats typical of Likert-scale questionnaires. The final model contained 15 items organized into two correlated factors, and its fit statistics were strikingly good: a chi-square of 165.45 with 89 degrees of freedom, a comparative fit index of 0.999, a Tucker-Lewis index of 0.999, a root mean square error of approximation of 0.065 with a 90 percent confidence interval of 0.049 to 0.080, and a standardized root mean square residual of just 0.038. In plain terms, the two-factor structure described the data almost perfectly.

But the story took an intriguing turn when the researchers compared this two-factor solution against a simpler one-factor alternative. The correlated two-factor model showed better descriptive fit than the single-factor model, suggesting that perseverance and passion are not identical. Both dimensions also demonstrated high internal consistency, meaning the items within each factor hung together reliably. Yet when the team examined the relationship between the two factors themselves, they found a latent correlation of 0.938 and a heterotrait-monotrait ratio, or HTMT, of 0.937. Both figures hover near the ceiling of what psychometricians consider acceptable for treating two constructs as genuinely distinct. In practical terms, the perseverance and passion factors were so tightly intertwined that the study could not establish discriminant validity, the demonstration that two supposedly separate traits are actually measuring different things.

The authors are refreshingly candid about this limitation. Their findings, they write, provide only preliminary support for a highly correlated two-factor measure, and they caution against interpreting the subscales separately or using them for high-stakes decisions such as hiring or promotion. They offer a compelling interpretation of the pattern: what looks like two factors may in fact reflect systemic discipline, a contextual pattern in which sustained effort and long-term commitment fuse into a single way of working within the structured environment of automotive manufacturing. In a factory, showing up consistently, maintaining precision, and committing to the long haul are not independent choices but interlocking habits reinforced by the production system itself. Grit in this setting may be less a matter of two separable psychological muscles and more a single, deeply embedded occupational disposition.

The implications reach well beyond Thailand’s automotive sector. For organizations, the study is a reminder that psychological constructs do not travel unchanged across contexts. A questionnaire validated on university students may quietly mismeasure the same trait in a welder, a nurse, or a logistics worker, and decisions based on such mis measurement can be costly and unfair. For researchers, the work illustrates the value of mixed methods scale development: qualitative interviews with experienced practitioners generated behavioral indicators no literature review alone would have produced, and quantitative modeling then tested whether those indicators formed a coherent, measurable structure. The two phases corrected each other, with the pilot data and confirmatory analysis trimming the initial 20-item pool down to a leaner, better-fitting 15-item instrument.

The study also contributes to a growing international conversation about the cultural boundaries of grit. Critics of grit research have long argued that the construct may reflect culturally specific ideals of individual striving, and that in collectivist settings, persistence is often expressed through loyalty to a team, a company, or a shared production goal. The Thai findings lend weight to that critique while offering a constructive path forward: rather than abandoning grit, researchers can rebuild its measurement from the ground up within each occupational and cultural context, letting the people who live the construct define its observable forms. The ten contextual indicators identified here, distilled from more than two centuries of combined supervisory experience, offer a template for how such reconstruction can proceed.

Much work remains before the new scale can be used with confidence. The authors explicitly call for independent replication, external validity testing against outcomes such as job performance or retention, and measurement invariance testing to confirm that the scale functions equivalently across different groups of workers. The sample of 204 employees, while adequate for confirmatory factor analysis, is drawn from a single industry in a single country, and the near-perfect correlation between factors means the two-factor structure may yet collapse into one. Still, the study stands as a model of methodological honesty: it reports stellar fit statistics and then immediately flags the discriminant validity problem that a less careful team might have buried. In a field often criticized for overclaiming, that combination of ambition and restraint may be the most durable form of grit on display.

Subject of Research: Development and psychometric validation of a context-specific grit scale for Thai automotive production workers

Article Title: A mixed methods development of a grit scale for Thai automotive operators

Article References: Artamaprasangsa, S., Tuntivivat, S., & Poonpol, P. (2026). A mixed methods development of a grit scale for Thai automotive operators. Discover Psychology. https://doi.org/10.1007/s44202-026-00934-x

Image Credits: AI Generated

DOI: 10.1007/s44202-026-00934-x

Keywords: grit, scale development, psychometrics, mixed methods, automotive operators, Thailand, confirmatory factor analysis, perseverance, occupational psychology, measurement invariance, manufacturing, survey methodology

The original GPS, conducted in 2012 as part of the Gallup World Poll, gathered responses from more than 80,000 people across 76 countries representing roughly 90 percent of the world’s population and economic output. It measures six experimentally validated preference domains that form the backbone of modern economic theory: risk-taking, patience, positive reciprocity, negative reciprocity, altruism and trust. These are not abstract curiosities. Patience predicts saving behaviour, educational attainment and entrepreneurial entry; trust correlates with financial development and growth; reciprocity underpins contract enforcement and workplace cooperation. If a language model could reproduce this preference landscape reliably, it could in principle generate counterfactual data for policy analysis, easing one of the deepest problems in causal inference — that we can never observe what would have happened under conditions that never occurred.

The research team, led by Fadong Chen and Qijun Wang of Zhejiang University together with Sudeep Bhatia of the University of Pennsylvania, took an unusually careful approach. Rather than asking models to answer surveys in the abstract, they built one-to-one ‘digital twins’: for each of 67,201 matched human respondents across 74 countries, they constructed a persona embedding that person’s age, gender, cognitive ability and country, then prompted each model to answer the identical GPS questionnaire. Crucially, the prompts were delivered in the respondents’ original survey languages, with payoffs denominated in local currencies — a Chinese respondent’s donation question, for instance, was framed in RMB. Twelve main model configurations, spanning OpenAI’s GPT family, Google’s Gemini models and DeepSeek-V3, generated roughly 27.8 million answers across about 1.39 million persona completions.

The headline finding is one of weak but genuine alignment. Pooling all countries, 63 of 72 correlations between model-predicted and human preferences — that is, 87.5 percent of the six preference domains across twelve configurations — were significantly positive. But the magnitude tells a different story. After normalizing both human and synthetic preference scores, the mean squared error between them averaged 1.827, nearly twice the variance of the human benchmark itself. Regression analyses showed that LLM-predicted preferences explained just 1.3 percent of the variance in individual human responses and, even after aggregating to country averages, only 4.2 percent of country-level variance. In plain terms, knowing what the AI ‘thinks’ a person would say tells you almost nothing about what that person actually said.

Perhaps the most striking pattern is that the models agree with each other far more than they agree with humans. Between-model mean squared errors were substantially lower than model-to-human errors at both the individual and country levels. This convergence suggests that today’s language models share a common output regularity — a kind of statistical accent — that systematically departs from the human benchmark. Notably, newer and larger models did not outperform older ones: GPT-5.4 showed no measurable advantage over GPT-3.5-turbo, and Gemini-2.5-Flash did not beat Gemini-2.0-Flash. Scaling alone, at least across these generations, does not appear to close the human-machine gap in economic preferences.

The audit then probed whether models reproduce the well-documented demographic gradients in preferences — differences by gender, age and cognitive ability. Here the results were more encouraging. Using a confidence-interval-based consistency criterion, models recovered 77.8 percent of the demographic patterns found in the GPS data, with cognitive-ability gradients matched in 84.7 percent of cases and gender gradients in 81.9 percent. Risk-taking and patience gradients were recovered perfectly across all twelve configurations. But cultural variation proved far harder. When the team tested whether models captured differences along fifteen cultural dimensions drawn from the World Values Survey, Hofstede’s framework and Schwartz’s value orientations, the consistency ratio collapsed to 48.1 percent — statistically indistinguishable from a coin flip. Only three of twelve models recovered the link between Hofstede’s long-term orientation and patience, and just one recovered the relationship between uncertainty avoidance and risk-taking.

Why do models handle demographics but stumble on culture? Ablation experiments, in which pieces of the persona prompt were systematically removed, provided a diagnostic answer. Stripping demographic information degraded alignment substantially, while removing the country label barely changed the results at all — individual-level error was 1.838 for full personas versus 1.961 when only country information was retained. The models, it seems, condition effectively on traits like age and gender but largely ignore the country tag, treating a persona in Nairobi much like one in Oslo. Demographic effect sizes in the ablation conditions fell from an average of 0.303 to nearly zero when demographic cues were absent, confirming that the recovered gradients depend on explicit prompt information rather than deep cultural knowledge.

The residual errors were not randomly distributed, either. The team constructed an ‘inconsistency index’ — the sampling-weighted within-country mean of individual prediction errors — and correlated it with national characteristics. In many models, prediction errors were significantly larger in countries with lower GDP per capita and weaker digital infrastructure, as measured by the ICT Development Index, while population size and internet-user counts showed no systematic relationship. The authors suggest two mechanisms: languages and content from wealthier countries receive disproportionate emphasis in model development, and digitized economies generate more of the high-quality structured data on which models train. The equity implications are uncomfortable — AI tools may simply work worse for the populations that stand to benefit most from careful policy analysis.

Practical replication tests drove the point home. When the researchers attempted to reproduce 35 published country-level findings from the GPS literature — such as the association between patience and GDP per capita, or gender gaps in preferences and gender equality — the best-performing model reproduced only 16, or 45.7 percent, with the same direction and statistical significance. Two additional biases emerged: women’s prosocial preferences were systematically overestimated across virtually all models, with the distortion largest in countries with stronger gender stereotypes, and model responses were far more concentrated than human ones, a homogenization that raising the sampling temperature only partially repaired.

The authors are careful not to declare LLM simulation useless. Synthetic respondents may still serve as tools for early-stage hypothesis generation, study-design diagnostics, robustness checks and ethically sensitive counterfactual exploration before human experiments begin, and the one-to-one digital-twin protocol itself offers a reusable benchmark for tracking how cultural alignment evolves across model generations. But the core conclusion stands as a warning against a seductive shortcut. Language models do not yet encode the cultural texture of human economic life, and their failures are patterned along lines of global inequality. Until that changes, understanding what real people want — in Nairobi, Oslo or anywhere else — still requires asking them.

Subject of Research: Validity of large language model-simulated survey responses as proxies for human economic preferences across 74 countries

Article Title: Simulated responses from large language models are weak proxies for human economic preferences across 74 countries

Article References: Chen, F., Wang, Q., Wang, L., & Bhatia, S. (2026). Simulated responses from large language models are weak proxies for human economic preferences across 74 countries. Nature Human Behaviour. https://doi.org/10.1038/s41562-026-02611-4

Image Credits: AI Generated

DOI: 10.1038/s41562-026-02611-4

Keywords: large language models, economic preferences, Global Preference Survey, survey methodology, cross-cultural alignment, digital twins, synthetic data, behavioral economics, AI bias, Nature Human Behaviour, policy evaluation, cultural diversity

HbA1c is a familiar number to anyone living with diabetes. It reflects the average blood glucose level over the preceding two to three months, because glucose binds irreversibly to haemoglobin in red blood cells in proportion to its concentration in the bloodstream. A value below 5.7 percent is considered normal, while 6.5 percent or above meets the diagnostic threshold for diabetes. The research team, led by Soo-Hyun Sung of Wonkwang University and corresponding authors Kwang-Il Park of Gyeongsang National University and Young Woo Kim of Dongguk University, used this marker together with treatment status to define four phenotypes. P1 describes people on no treatment or lifestyle modification alone whose HbA1c remains below 7 percent, the conventional target for glycaemic control. P2 captures those who have reached their target while taking oral glucose-lowering medication. P3 covers people with HbA1c between 7 and 9 percent who are not yet on insulin, and P4 is the most severe category, comprising anyone using insulin or whose HbA1c has risen to 9 percent or higher.

The dataset behind the analysis is substantial. From the survey years 2015 through 2024, the researchers identified 8,702 adults aged 30 or older who met at least one of four criteria for diabetes: a physician diagnosis, current use of glucose-lowering medication, a fasting glucose of 126 milligrams per decilitre or more, or an HbA1c of 6.5 percent or above, combined with a measured HbA1c value. Because KNHANES uses a complex, stratified, multistage sampling design, the team applied survey weights and design-based statistical methods throughout, ensuring that their estimates reflect the Korean adult population rather than the quirks of the sample. The weighted distribution across the four phenotypes came out as 24.8 percent in P1, 34.3 percent in P2, 28.9 percent in P3, and 12.0 percent in P4.

The headline finding is not the distribution itself but its sudden rearrangement. A formal change-point analysis, which tests whether a time series is better explained by a single trend or by two segments separated by an abrupt shift, selected 2022 as the break point for both the P2 and P3 phenotypes. The selection was not marginal. The year 2022 was chosen by both the design-based Akaike information criterion, a statistic that penalises model complexity, and the maximum Wald statistic, and it reappeared in 97.1 percent and 97.7 percent of 1,000 design-based bootstrap replicates for P2 and P3 respectively. Critically, a discrete level shift fitted the data better than a gradual change in slope, suggesting a step change rather than a slow drift in the composition of the diabetes population.

The magnitude of the shift is striking. The share of people in the treated, at-target phenotype P2 rose from 28.2 percent during 2015 to 2021 to 46.8 percent during 2022 to 2024, a relative increase of 66 percent, equivalent to 18.6 percentage points. After adjustment for covariates, the prevalence ratio for belonging to P2 in the later period was 1.49, with a 95 percent confidence interval of 1.31 to 1.69. Over the same interval, P1 and P3 both declined, with adjusted prevalence ratios of 0.81 and 0.68 respectively, both statistically significant, while P4, the insulin-treated or severely hyperglycaemic group, remained essentially unchanged. In practical terms, the population appears to have migrated from the untreated and above-target categories into the well-controlled, orally treated category, while the most severe group held steady.

Here the study delivers its most important caveat. The year 2022 is also the first year of KNHANES cycle IX, a redesign of the survey program. The authors are explicit that the observed discontinuity cannot be attributed to genuine changes in diabetes care within this study design, because a change in survey methodology, sampling frame, or measurement procedures would produce exactly the kind of step change they detected. This is a recurring challenge in repeated cross-sectional surveillance: when the instrument changes, the signal and the artefact become difficult to disentangle. The researchers therefore frame their temporal finding as a description of a redistribution in the data, one whose clinical interpretation requires confirmation from other data sources and survey designs.

The analysis of correlates, the factors associated with belonging to one phenotype rather than another, also carries a methodological twist. Because the reference category P1 is dominated by screen-detected cases, 88 percent of its members had never been previously diagnosed, comparisons against it can generate associations that reflect diagnostic status rather than biology. When the researchers restricted the analysis to previously diagnosed participants, most of the associations seen in the full sample shrank or vanished. Reduced kidney function, defined as an estimated glomerular filtration rate below 60, was the only association with the insulin-treated or markedly hyperglycaemic phenotype that retained its full magnitude, with an odds ratio of 2.67 and a confidence interval of 1.27 to 5.62. This makes physiological sense: declining renal function is both a complication of long-standing diabetes and a factor that shapes treatment decisions, often pushing clinicians toward insulin-based regimens.

Reported dyslipidaemia, the presence of abnormal blood lipid levels, persisted as a correlate of the P2 and P3 phenotypes but at roughly half its full-sample strength, with an odds ratio of 1.79 in the diagnosed-only analysis. Meanwhile, the inverse associations with obesity and alcohol consumption that appeared in the full sample disappeared entirely once the comparison was restricted to diagnosed individuals, a pattern the authors attribute to the composition of the screen-detected reference group. In the full sample, reported descriptively because of this reference-group problem, the corresponding estimates for reduced kidney function were considerably larger, at 3.09 for both comparisons. The lesson is a general one for observational diabetes research: who ends up in your comparison group can manufacture or erase associations, and screening status is a powerful hidden variable.

The robustness work behind the paper is unusually thorough for a survey analysis. A sensitivity analysis using multiple imputation, a statistical technique that fills in missing covariate values by drawing from their predicted distributions, reproduced every complete-case estimate closely, with the largest discrepancy amounting to just 0.17 on the odds-ratio scale. The temporal findings survived three alternative classification rules, and the estimated change-point remained 2022 under each of them. Of 42 interaction tests examining whether the associations differed across subgroups, none remained significant after correction for the false discovery rate, the expected proportion of false positives among declared findings. These checks do not eliminate the survey-redesign confound, but they do indicate that the reported patterns are not artefacts of missing data or arbitrary analytical choices.

The authors are careful about what their classification does and does not claim. The four phenotypes describe where people currently sit on the combined axes of glycaemic control and treatment intensity; they are not a prognostically validated staging system, and the scheme’s ability to predict complications, cardiovascular events, or mortality remains to be tested in longitudinal and independent populations. Still, the study offers a template for how national surveillance data can be mined more intelligently. By replacing the blunt controlled-versus-uncontrolled dichotomy with a composite phenotype that respects the heterogeneity of the diabetes population, and by subjecting temporal trends to formal change-point analysis with bootstrap validation, the researchers have shown both the promise and the peril of such approaches. The 2022 discontinuity in Korean diabetes data may eventually prove to reflect real changes in care, new drug classes such as SGLT2 inhibitors and GLP-1 receptor agonists reshaping treatment patterns, or simply the fingerprint of a redesigned survey. Until that question is resolved with independent evidence, the study stands as a rigorous description of a population in flux, and a reminder that in health statistics, the measuring instrument is itself part of the measurement.

Subject of Research: Glycaemic–treatment phenotypes of diabetes and their temporal change among Korean adults

Article Title: Glycaemic–treatment phenotypes of diabetes and their correlates among Korean adults: a repeated cross-sectional analysis of the Korea National Health and Nutrition Examination Survey, 2015–2024

Article References: Sung, S.-H., Lee, J., Park, H.-G., Lee, J.-H., Sung, H.-K., Lee, S.-U., Yang, J.-H., Park, K.-I., & Kim, Y. W. (2026). Glycaemic–treatment phenotypes of diabetes and their correlates among Korean adults: a repeated cross-sectional analysis of the Korea National Health and Nutrition Examination Survey, 2015–2024. BMC Endocrine Disorders. https://doi.org/10.1186/s12902-026-02573-8

Image Credits: AI Generated

DOI: 10.1186/s12902-026-02573-8

Keywords: diabetes, HbA1c, KNHANES, phenotype classification, change-point analysis, glycaemic control, insulin treatment, chronic kidney disease, dyslipidaemia, South Korea, survey methodology, epidemiology

The stakes are higher than many researchers realize. Previous work cited in the study shows that a careless responding rate of roughly 14 percent inflated correlations in published psychological studies by about .07, while simulations have demonstrated that even 5 percent careless responding can produce sign reversals in estimated correlations. Other research has found that removing as little as 4 to 10 percent of careless responders improved model fit in empirical datasets, and that just 10 percent contamination can substantially bias estimates of item discrimination parameters. Because researchers never know in advance how contaminated their data are, the authors argue, evaluating results with and without adjustments for presumed carelessness is essential for trustworthy science.

Traditionally, researchers have hunted for careless responders using screening indices: the long-string index that counts how many identical answers appear in a row, the intraindividual response variability index, the even–odd consistency index that correlates halves of a scale, Mahalanobis distance for multivariate outliers, and attention check items planted in the questionnaire. The trouble is that each index catches only one flavor of carelessness, so researchers typically stack several of them into a multihurdle approach. Worse, every hurdle requires a threshold—a cutoff above which a respondent is declared careless—and there are no firmly grounded rules for setting those thresholds. Small changes in cutoff choices have repeatedly been shown to produce substantially different classification rates, handing researchers enormous and often invisible degrees of freedom.

Mixture models offer an alternative that sidesteps the threshold problem entirely. The idea is elegant: assume that the sample contains two latent groups, one of attentive respondents whose answers reflect the traits being measured, and one of careless respondents whose answers do not. A standard measurement model—in this study, a multidimensional graded response model appropriate for categorical Likert items—links responses to latent traits in the attentive class. In the careless class, the model assumes responses do not depend on item content at all: threshold parameters are constrained to be equal across items, and a single latent variable representing each respondent’s preference for certain response categories replaces the substantive traits. The model then estimates, for every person, a posterior probability of belonging to the attentive class, conveying not just a verdict but a degree of certainty. A posterior probability of .60 signals that the response pattern says little either way; a .90 signals strong evidence of attentiveness.

The exemplar model used in the study was deliberately designed to capture the core features shared by most existing mixture approaches for C/IER while avoiding their more restrictive assumptions. Unlike many predecessors that treat item responses as continuous, it acknowledges their categorical nature, allowing the shape of response distributions to differ between classes—for instance, roughly normal for attentive respondents versus uniform for random responders. It also accommodates dependencies among careless responses arising from individual differences in category preferences. Because the specification retains the two-class structure common to the wider family of mixture models, the authors expect their findings to generalize to closely related approaches.

The heart of the study is a simulation comprising 240 crossed conditions, each replicated 200 times, with samples of 500 respondents and 10 percent careless responding—a realistic baseline given reported rates of roughly 5 to 12 percent in offline studies and 6 to 10 percent in large-scale assessments, but also an extreme class imbalance that makes detection genuinely hard. The researchers varied the number of five-point Likert items (5, 10, or 20), the proportion of negatively worded items (0 or 40 percent), the heterogeneity of item thresholds, the strength of item discriminations, and the type of careless behavior: straightlining, respondent-specific category preference, random responding, or a blend of the last two. Three criteria were evaluated: whether model comparisons correctly detected the presence of C/IER using the Bayesian information criterion, whether the estimated proportion of careless respondents was accurate, and whether individuals were correctly classified.

The results reveal a clear recipe for success. When the data contained no careless responders at all, the simpler model was always correctly preferred—no false alarms. When 10 percent careless responding was present, straightlining was detected consistently under every condition tested. For the subtler patterns, however, questionnaire design mattered enormously. With only five items, reliable detection generally required scales mixing positively and negatively worded items together with high item heterogeneity, meaning items differ substantially in how easy they are to endorse. With ten items, either mixed wording or high heterogeneity sufficed. With twenty items, the mixture model was preferred in nearly every condition regardless of wording or heterogeneity. The logic is intuitive: on a short, uniformly worded scale of easy-to-endorse items, an attentive respondent who consistently picks the category that best describes them produces a pattern nearly indistinguishable from a straightliner. Paradoxically, stronger item discriminations can make this problem worse, sharpening the tendency of attentive respondents to answer uniformly, while under random responding the effect reverses and weak discriminations become the hazard.

Individual-level classification told a similar story. Specificity—the probability of correctly assigning simulated attentive respondents to the attentive class—exceeded .90 in every single condition, partly because with only 10 percent contamination, a classifier defaulting to attentive already achieves 90 percent specificity. Sensitivity, the ability to catch true careless responders, depended heavily on scale length, item heterogeneity, and wording, rising with all three for most careless response types. For estimating the overall proportion of careless respondents, mixed-worded scales with high heterogeneity kept bias within acceptable bounds even at five items, while unidirectional, low-heterogeneity, short scales inflated estimates toward 14 percent. Straightlining proved uniquely stubborn: the proportion of careless respondents was systematically underestimated, dropping as low as 6 percent, and longer scales actually made the underestimation worse rather than better.

To show that the latent classes really mean what researchers hope they mean, the team reanalyzed a publicly available Big Five dataset collected across four online crowdsourcing platforms—MTurk, CloudResearch, Prolific, and Qualtrics—with roughly 500 respondents each, using the 50-item NEO-PI-R inventory in which each trait is measured by ten items, half negatively worded. By the simulation’s own criteria, these scales offered favorable conditions for detection. The model classified between 76.5 and 83.3 percent of respondents as attentive depending on the scale, and classifications were strikingly consistent: 62.4 percent of respondents were classified as attentive on all five scales, while only 8.4 percent failed to reach the attentive class on any scale, with posterior probabilities across scales correlating between .509 and .666. This cross-scale stability suggests the model captures a stable disposition toward careless responding rather than scale-specific quirks.

Crucially, the model-based classifications converged with evidence that does not rely on response patterns at all. Respondents who passed more of the five embedded attention check items received higher posterior probabilities of being attentive, with virtually all respondents passing four or five checks classified as attentive with near-certainty across all scales. Even more striking, the latent class variable tracked the data-collection platform more strongly than the attention check index did: averaged across scales, Nagelkerke’s pseudo-R-squared was 0.224 for the model-based classification versus 0.071 for the index. Both methods identified MTurk as by far the worst platform, but the model painted a starker picture—only 47 to 59 percent of MTurk respondents were classified as attentive on some scales, compared with 85.6 percent flagged by attention checks, while Prolific fared best at 88 to 95 percent attentive. The authors conclude that with scales of at least ten items, mixed wording, and sufficient item heterogeneity, mixture modeling is a viable, comparatively easy-to-implement tool for addressing careless responding—available in standard software such as the R package mirt—while cautioning that short, uniformly worded, low-heterogeneity scales remain its blind spot, and that questions about trait-score accuracy and misclassification consequences await future research.

Subject of Research: Detection of careless and insufficient effort responding in self-report questionnaires using latent variable mixture models

Article Title: Understanding when mixture models effectively identify careless responding: Simulation and validity evidence

Article References: Uglanova, I., Nagy, G., & Ulitzsch, E. (2026). Understanding when mixture models effectively identify careless responding: Simulation and validity evidence. Behavior Research Methods, 58(11), Article 313. https://doi.org/10.3758/s13428-026-03165-z

Image Credits: AI Generated

DOI: 10.3758/s13428-026-03165-z

Keywords: careless responding, mixture models, item response theory, survey methodology, psychometrics, latent class analysis, attention checks, straightlining, Big Five, data quality, online platforms, simulation study

The research, published in the Journal of Environmental Studies and Sciences by a team led by Emma Hankinson of Oxford Brookes University alongside experienced consultant ecologists, evaluated 38 commercial reptile survey reports produced between 2009 and 2024. The reports covered three of Britain’s lesser-protected native reptiles: the common lizard (Zootoca vivipara), the slow worm (Anguis fragilis) and the grass snake (Natrix helvetica). All had been drafted to inform planning applications under the Town and Country Planning Act 1990, and all were publicly available online under the Environmental Information Regulations 2004.

To measure quality, the team built a novel nine-category scoring rubric drawn from both statutory guidance and the wider best-practice literature, including the Herpetofauna Workers Manual, Froglife advice sheets and peer-reviewed survey protocols. Categories ranged from habitat suitability assessments and preliminary data searches to weather conditions, survey seasonality, the number of presence or absence surveys, population size class estimates, use of artificial cover objects and transects, surveyor experience and recommended monitoring. Each report was rated on a four-point ordinal scale per category, yielding a total score between 9 and 36. A second independent rater scored every report, and agreement was strikingly high, with a pooled weighted Cohen’s kappa of 0.98 and an intraclass correlation of 0.99 on total scores, confirming the rubric could be applied consistently.

The headline finding is a stark split between legal compliance and scientific best practice. Reports scored significantly higher on categories anchored in statutory guidance, with a mean item score of 3.09, than on non-statutory best-practice categories, which averaged just 1.95, a difference confirmed by a Wilcoxon signed-rank test at p < 0.001. In other words, consultants reliably tick the boxes the law requires but routinely skip the techniques that would make their surveys genuinely robust. The weakest categories of all were population size class estimates and recommended monitoring, each averaging a dismal 1.13 out of 4. Although every report confirming reptile presence provided some population estimate, more than three quarters based it solely on the initial presence or absence surveys without any additional effort. Even more striking, 85.71 percent of reports mentioned no post-construction monitoring whatsoever, meaning the effectiveness of mitigation is almost never verified.

This pattern has deep roots. Britain’s current reptile survey methodologies rest on guidance more than two decades old, and earlier attempts to modernise it faltered; Natural England updated its reptile guidance in 2011 only to withdraw it within the same year over concerns about the survey effort it demanded. More recent documents tend to reference older guidance circularly rather than incorporating newer, evidence-based methods. Because reptiles are small, cryptic and highly seasonal, survey design matters enormously, and research consistently shows that greater effort and varied techniques dramatically improve detection. Yet the statutory floor remains low, particularly for the four widespread species, whose habitats enjoy no formal legal protection at all.

Perhaps the most consequential finding concerns reports that conclude reptiles are absent. Confirming absence reliably requires far more survey effort across a wider seasonal window than confirming presence, yet absence reports scored significantly lower on precisely the two categories most tied to that evidentiary standard: survey month and season (p = 0.002) and the number of presence or absence surveys undertaken (p = 0.022). Nearly 41 percent of absence reports conducted all their surveys in sub-optimal months, and almost 18 percent undertook no surveys at all, simply declaring the habitat unsuitable. Grass snakes, for example, linger for weeks around egg-laying sites in June and July; if those months go unsampled or the sites fall outside the survey area, the species can be missed entirely. A false absence conclusion is legally corrosive, because it removes any statutory obligation to mitigate harm, whereas a correctly identified presence triggers protection.

There are faint signs of progress. Total rubric scores rose significantly across the 2009 to 2024 period, with a Spearman correlation of 0.378, despite no change in statutory guidance, suggesting that evolving professional norms, sector-wide guidance from bodies such as CIEEM and increased scrutiny of environmental compliance are nudging standards upward. Report quality was also positively associated with the severity of development impact reported, though intriguingly, rubric score showed no relationship at all with the mitigation strategies recommended. The dominant mitigation was translocation of reptiles to non-adjacent habitat they had not previously occupied, an approach with a troubling evidence base: one study recorded only a 1.6 percent recapture rate of translocated reptiles during subsequent monitoring, and other work has documented displaced lizards in poor body condition with elevated parasite loads.

The authors argue that statutory guidance alone is insufficient to guarantee reliable survey outcomes, and they point to an alternative delivery model already proving itself for amphibians. The NatureSpace district licensing scheme for great crested newts replaces site-by-site consultant surveys with a single centrally funded body that creates and manages compensatory habitat strategically, in advance of development, backed by 25-year monitoring agreements. Published results from 2021 to 2025 show 90 percent of mature compensation sites already occupied by target and other species. With the Planning and Infrastructure Act 2025 establishing Environmental Delivery Plans and a Nature Restoration Fund along similar lines, reptiles could be strong candidates for such a strategic approach, though the authors caution that any extension would need robust monitoring safeguards, since some conservation organisations fear these schemes may weaken project-level scrutiny.

The study has honest limitations. Reports were found via a single search engine, likely favouring consultancies with stronger online visibility; one report per consultancy means results are unweighted by output volume; and the rubric, while internally reliable, was devised by the authors and has not been externally validated, with its quality thresholds remaining partly subjective. The modest sample of 38 reports also limits statistical power for subgroup comparisons. Crucially, the analysis could only assess what consultants self-reported, with no independent means of verifying the underlying data.

Even so, the picture that emerges is sobering for a country now ranked among the most nature-depleted on Earth. Reptiles are exceptionally sensitive to disturbance: they cannot recolonise across long distances, cannot wait out adverse conditions, and are acutely vulnerable to small climatic shifts. Britain’s reptiles were first recognised as declining in the 1970s and 1980s, and recent assessments suggest nearly a third of the nation’s amphibian and reptile species face extinction risk. If the reports that decide their fate meet only the legal minimum, and the legal minimum is anchored to outdated, quasi-quantitative methods, then countless small failures may be compounding into a landscape-scale problem. The authors’ prescription is clear: revise statutory guidance to embed monitoring, habitat suitability assessment and adequate survey effort, and seriously consider whether centrally delivered, long-term compensated conservation, already working for newts, should now extend to the snakes and lizards slipping through the current system’s cracks.

Subject of Research: Quality and statutory compliance of commercial reptile survey reports informing development planning in Great Britain

Article Title: Falling short: inconsistent statutory compliance and widespread best-practice neglect in commercial reptile survey reports across Great Britain

Article References: Hankinson, E., Bartlett, K., Goldsmith, J., Sharma-Vicary, K., & Nijman, V. (2026). Falling short: inconsistent statutory compliance and widespread best-practice neglect in commercial reptile survey reports across Great Britain. Journal of Environmental Studies and Sciences. https://doi.org/10.1007/s13412-026-01148-2

Image Credits: AI Generated

DOI: 10.1007/s13412-026-01148-2

Keywords: reptile conservation, ecological consultancy, survey methodology, statutory guidance, mitigation translocation, great crested newts, biodiversity net gain, planning policy, herpetofauna, Great Britain, wildlife protection, monitoring

Discrete choice experiments rest on a deceptively simple premise: rather than asking people what they want in the abstract, researchers present them with a series of hypothetical scenarios, each describing a healthcare provider defined by a handful of characteristics, and ask which they would prefer. By varying those characteristics systematically across many choices, statisticians can estimate the relative weight that respondents place on each one, and even calculate how much of one attribute, such as money, a person would sacrifice to gain an improvement in another, such as shorter waiting time. The technique has become a mainstay of health economics in high-income countries, but its application in low- and middle-income settings has been constrained by a persistent problem: the attributes, the characteristics that define the choices, are often borrowed from studies conducted in entirely different contexts, where the drivers of healthcare decisions may be radically different.

That is the gap the new study set out to close. Rapid urbanisation across South Asia and much of the developing world has outpaced the growth of urban primary healthcare systems, and in cities like Dhaka the slum population relies on a bewildering patchwork of providers, from government clinics and non-governmental organisation facilities to private pharmacies, informal drug sellers and traditional practitioners. Despite the scale of this challenge, remarkably little research has asked the residents of these settlements what they actually value when choosing among such providers. Without that knowledge, policymakers designing urban health services are effectively guessing at the preferences of the very people they aim to serve, and the risk is that new facilities or financing schemes fail to attract patients because they overlook the features that matter most on the ground.

The research team approached the problem in two carefully sequenced phases designed to be transparent and replicable by others working in similar settings. The first phase was about breadth: casting a wide net to capture every plausible attribute of primary healthcare that might influence a slum dweller’s choice. The investigators drew on three complementary sources. They conducted a systematic review of the relevant literature to identify attributes used in previous healthcare choice studies. They ran four focus group discussions with community members in Dhaka’s slums, deliberately including both people who had used primary healthcare services in the previous year and those who had not, to capture the perspectives of users and non-users alike. And they carried out eight key informant interviews with supply-side stakeholders, the people who plan, fund and deliver urban health services, to understand the considerations from the provider perspective as well.

This triangulation of evidence produced a long list of twenty-five distinct attributes, ranging from clinical quality and staff behaviour to distance, cost and social recommendation. A list of that length is unmanageable for a discrete choice experiment: respondents presented with too many characteristics at once become overwhelmed, and the statistical burden of estimating preferences for dozens of attributes grows rapidly. The second phase of the study was therefore about reduction and refinement. The team narrowed the long list to twelve attributes and took them to the community itself, conducting a ranking exercise with fifteen slum residents who were asked to order the attributes by importance. They then consulted one-to-one with four urban health experts, who scrutinised the shortlist for feasibility, policy relevance and interpretability, and finally pretested the draft survey instrument with members of the target population to check that the attributes and their levels were actually understood as intended.

Out of this iterative process emerged six attributes for the final experiment: the travel time required to reach a healthcare centre, the waiting time once there, the type of healthcare provider, whether the provider had been recommended by neighbours or relatives, the consultation fee, and the expenditure on medicines. The selection is striking for what it reveals about healthcare decision-making in this setting. Two of the six attributes are purely temporal, capturing the total time burden of care from doorstep to consultation room. Two are financial, separating the visible, negotiated cost of the consultation from the often larger and less predictable cost of medicines, a distinction that matters enormously for households where out-of-pocket payments dominate health spending. And two are social and informational: the identity of the provider and the word-of-mouth endorsement of trusted community members, which in neighbourhoods where formal quality signals are scarce may serve as the primary heuristic for judging competence and trustworthiness.

The attribute levels, the specific values each characteristic can take in the hypothetical scenarios, were grounded directly in the qualitative data rather than in administrative records or expert assumptions. The range of travel times, fees and medicine costs reflected the real experiences described by focus group participants, ensuring that the choices presented in the eventual survey would be recognisable and plausible to respondents. This grounding in lived experience is one of the study’s most important methodological contributions, because attribute levels that fall outside the range of what people have actually encountered can distort preference estimates, producing results that are statistically tidy but practically meaningless for policy.

Perhaps the most instructive findings came during pretesting, when the researchers confronted a barrier that is easy to underestimate: literacy. Comprehension of the attributes and levels proved difficult for participants with low literacy, a common characteristic in slum populations where formal schooling may be limited or interrupted. The team responded by revising the wording of the survey and introducing graphical Q-cards, visual representations of each attribute and level that allow respondents to grasp the choices without relying on text. The lesson has broad implications for survey research in low-income urban settings worldwide, where instruments designed for educated populations can silently exclude or mislead the very respondents whose preferences are most needed. A choice experiment whose scenarios are misunderstood does not measure preferences; it measures confusion.

The study, which received ethics approval from the University of Leeds and BRAC University and was funded through the CHORUS Research Programme Consortium supported by the UK Foreign, Commonwealth and Development Office, is candid about its own scope: it does not report the preference results themselves, which await the full discrete choice experiment. What it delivers instead is something arguably more durable, a documented, step-by-step template for building choice experiments from the ground up in understudied, low-income urban environments. By showing how qualitative evidence, community ranking, expert consultation and iterative pretesting can be woven together to distil twenty-five candidate attributes down to six defensible ones, the authors offer researchers across the global South a practical worked example that resists the temptation to import attribute sets from unrelated contexts.

For the residents of Dhaka’s slums, the eventual payoff could be concrete. Once the choice experiment is fielded, its results will allow health planners to estimate, for example, how much a reduction in waiting time would increase uptake of a government clinic, or how sensitive patients are to consultation fees relative to medicine costs, evidence that can be used to design services people will actually use. In a world where more than a billion people live in informal settlements and urban health systems are straining to keep pace, understanding the arithmetic of healthcare choice at the household level is no longer an academic luxury. This study demonstrates that the tools of health economics can be adapted, with care and humility, to hear what the urban poor are actually saying about the care they want, and it hands the next generation of researchers a rigorous map for following the same path.

Subject of Research: Development of attributes and attribute-levels for a discrete choice experiment on primary healthcare provider preferences among urban slum dwellers in Bangladesh

Article Title: Developing attributes and attribute-levels for a discrete choice experiment on preference for primary healthcare providers among urban slum dwellers in Bangladesh

Article References: Hasan, M. Z., Webb, E. J., Tisha, K. I., Quayyum, Z., & Ensor, T. (2026). Developing attributes and attribute-levels for a discrete choice experiment on preference for primary healthcare providers among urban slum dwellers in Bangladesh. BMC Public Health. https://doi.org/10.1186/s12889-026-29585-2

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29585-2

Keywords: discrete choice experiment, primary healthcare, urban slums, Bangladesh, Dhaka, health economics, attribute development, healthcare preferences, low- and middle-income countries, survey methodology, out-of-pocket payments, urban health

The study’s design was straightforward but methodologically careful. The researchers collected responses from 97 third-year nursing students across two campuses of the University of Santiago de Compostela, in Santiago de Compostela and Lugo. They then administered a selection of items from the Spanish Health Barometer to this student group and compared the resulting answers with secondary data from 411 young adults aged 18 to 25 who had participated in the 2024 edition of the national survey. This created a natural comparison between two populations of similar age, one embedded in professional training and one drawn from the general public. The descriptive comparative design was applied across both sites of the single university, and ethical approval for the anonymised analysis of student responses was granted by the university’s Research Ethics Committee in January 2025, with all participants providing informed consent before completing the questionnaire.

The statistical treatment deserves attention because it reflects a growing trend toward rigor in educational research. Rather than relying on raw significance tests alone, the authors employed an omnibus chi-squared test for categorical comparisons and Welch’s t-tests for continuous measures, the latter being a variant that does not assume equal variances between groups and is therefore more robust when sample sizes and spreads differ. Crucially, they reported effect sizes using Hedges’ g, a measure that expresses the magnitude of a difference in standardized units, and they adjusted their p-values using the Benjamini–Hochberg procedure, which controls the false discovery rate when many comparisons are made simultaneously. This combination means the reported differences are not merely statistical artifacts of multiple testing; they survived a correction designed to filter out spurious findings.

The headline numerical result was that students rated the overall functioning of the health system at a mean of 7.00 on the survey’s scale, while young adults in the national dataset averaged 6.47. The difference was highly significant, with a p-value below 0.001 and an effect size of g = 0.326, which conventionally falls in the small-to-moderate range. On its face, this suggests that nursing students hold a somewhat more favorable view of the system’s overall performance than their peers outside health professional training. One plausible interpretation is that immersion in clinical placements and coursework gives students a more textured, less media-mediated picture of how the system operates day to day, including its successes as well as its strains.

Yet the categorical item told a different and more intriguing story. When asked to characterize how the system works in broader terms, only 4.1 percent of students judged that it worked quite well, compared with 11.0 percent of young adults in the national sample. Meanwhile, 40.2 percent of students said fundamental changes were required, against 31.1 percent of the general young adult population. In other words, the same students who awarded the system a higher average score were simultaneously more likely to call for deep structural reform. This divergence between a continuous rating and a categorical judgment is a textbook illustration of why survey methodology matters: the format of a question can shape, or at least reframe, the attitudes it captures, and researchers who read only one item risk drawing incomplete conclusions about public and professional sentiment.

The item-level comparisons deepened the pattern. Students rated hospital emergency services significantly higher than their young adult counterparts, with the difference again significant at p < 0.001 and an effect size of g = 0.312. This enthusiasm for emergency care may reflect direct exposure to hospital departments during clinical rotations, where students witness the machinery of acute medicine at close range. By contrast, students expressed significantly lower trust in primary care physicians, with an effect size of g = −0.318, and gave significantly lower ratings to the time that physicians dedicate to patients, at g = −0.423, the largest effect recorded in the study. Both findings remained significant after the Benjamini–Hochberg correction. One of the most notable null results concerned nursing itself: trust in nurses did not differ significantly between the two groups, suggesting that whatever professional socialization occurs during a nursing degree, it does not measurably inflate students’ confidence in their own future profession relative to the public’s view.

Why would nursing students be more critical of primary care than the general public? The study does not definitively answer this question, but the educational context offers reasonable hypotheses that the authors frame carefully. Students in their third year have typically completed rotations that include community and primary care settings, where they may observe time pressures, administrative burdens, and continuity-of-care challenges from an insider’s vantage point. Their lower ratings of the time physicians dedicate to patients could reflect an acute awareness of consultation length as a determinant of care quality, a theme heavily emphasized in nursing curricula that foreground holistic, patient-centered practice. At the same time, the authors are explicit that these are perceptions, not objective measures of system performance, and that the educational application arising from the study requires formal implementation and evaluation before any causal claims can be made.

The larger significance of the work lies in its proposal to treat national health surveys as pedagogical instruments. Surveys such as the Spanish Health Barometer, along with international counterparts like the European Health Interview Survey, the United States Behavioral Risk Factor Surveillance System, and national instruments in other countries, generate rich, publicly available datasets that policymakers rely on routinely. Yet, as the authors note, these resources remain underexplored in nursing education as tools for fostering critical thinking and systems-level awareness. The approach tested here inverts the usual classroom dynamic: instead of students merely answering survey questions, they compare their own answers against a nationally representative benchmark, forcing them to confront where their views align with the public and where they diverge, and to reason about why. That exercise sits at the intersection of survey methodology, health policy literacy, and professional identity formation, three competencies that modern nursing programs increasingly aim to cultivate together.

The study’s limitations are worth stating plainly, and the authors do so. The student sample came from a single university across two campuses, with 97 participants, which limits generalizability to nursing students in other regions or countries. The comparison group was restricted to young adults aged 18 to 25 within the national survey, so the findings speak to generational perceptions rather than the full population. The questionnaire had originally been administered as a voluntary and anonymous educational activity, and the research use involved secondary analysis of anonymised responses, a design that is ethically efficient but constrains the depth of data collection. And because the design is descriptive and comparative rather than experimental, it can document differences in perception but cannot explain their causes or test whether exposure to survey data actually improves educational outcomes.

Those caveats aside, the study opens a concrete agenda for future research. The authors call for intervention studies that formally evaluate whether the structured use of national survey data in nursing curricula supports critical thinking, systems literacy, and policy awareness. If such studies succeed, the implications could extend well beyond Spain. Health ministries and statistical agencies worldwide already invest heavily in opinion and health surveys; harnessing those same datasets as mirrors for health professional students would cost little and could reshape how the next generation of nurses understands the systems they will staff, defend, and reform. In an era when health systems everywhere face workforce shortages, aging populations, and eroding public trust, teaching future nurses to read the public’s verdict on their own institutions, and to compare it honestly with their own, may prove to be one of the most quietly consequential innovations in nursing education. This study, modest in size but rigorous in method, offers the first empirical foothold for that idea.

Subject of Research: Comparing nursing students' and young adults' perceptions of the Spanish health system using national survey data for nursing education

Article Title: National health survey data as a resource for nursing education: a comparative study of nursing students’ and young adults’ perceptions of the health system

Article References: Díaz, M. J. F., & Tuñas, M. C. A. (2026). National health survey data as a resource for nursing education: a comparative study of nursing students’ and young adults’ perceptions of the health system. BMC Nursing. https://doi.org/10.1186/s12912-026-05462-6

Image Credits: AI Generated

DOI: 10.1186/s12912-026-05462-6

Keywords: nursing education, Spanish Health Barometer, health system perceptions, survey methodology, critical thinking, primary care, comparative study, health policy, nursing students, effect sizes, public trust, systems literacy