Unilateral spatial neglect, or hemineglect, is a disabling cognitive consequence of right-hemisphere stroke in which a person fails to orient toward or interact with stimuli in the contralesional, usually left, side of space despite having intact primary sensory and motor function. The condition can interfere with reaching, object manipulation and self-care, and it is associated with longer hospitalization and poorer long-term outcomes. HemiBot is intended for post-stroke neglect through a non-contact exercise: the end effector of PAL Robotics’ TIAGo mobile manipulator acts as a moving visual cue while a participant reaches toward it with the affected arm. The robot does not physically guide the limb, and the task is designed to combine attentional cueing with active movement into the neglected hemispace.

The underlying software is divided into two cooperating layers. A session-level Coach node loads a YAML session plan, arranges exercises, receives aggregate results, presents a real-time scoreboard to the therapist and selects the difficulty for a subsequent session. At the individual-repetition level, a Python Behavior Tree built with py_trees coordinates motion commands, skeletal tracking, metric calculation, session aggregation and emergency-stop handling. This modular arrangement allows components such as the hand-pointing metric to be replaced later without redesigning the overall orchestration. After each attempt, the robot also delivers a text-to-speech message in Italian chosen from predefined pools associated with acceptable, failed or high-quality performance.

For the pointing exercise, the TIAGo gripper travels along a semicircular arc centered on the participant’s left shoulder. Easy, medium and hard settings use arc extents of 20, 30 and 40 degrees, respectively, while sharing the same 0.85-meter radius, angular velocity, three repetitions per standard session and dwell time at the extremes. The end effector always faces the arc center. Its movement is generated through a custom C++ ROS 2 action server using MoveIt 2 Cartesian path planning, with a trapezoidal velocity profile applied to the joint trajectory. Limits on joint velocity and acceleration are checked before execution, and invalid trajectories abort the action. Progress feedback is published at 5 hertz, while the Behavior Tree’s main tick runs at 30 hertz.

The system also treats the participant as a moving obstacle. Skeletal joint estimates from the robot’s onboard RGB-D camera are converted into spherical collision objects and updated in the MoveIt 2 planning scene. An OctoMap built directly from the depth stream serves as a coarser fallback when skeletal tracking temporarily deteriorates. This combination is intended to keep the planned trajectory away from the participant’s current body configuration, although the reaching targets themselves are designed a priori to remain at a safe distance. Emergency-stop signals are handled both by the arm action server and the Behavior Tree; the software detects the command within approximately 33 milliseconds at its 30-hertz control rate, but actuator delay, physical stopping time and stopping distance were not measured. The authors therefore describe the safety mechanisms as design provisions requiring quantitative validation before clinical use.

Movement quality is estimated from two quantities captured at 30 hertz: the angle between the line from the participant’s shoulder to the wrist and the line from the shoulder to the robot’s end effector, and the Euclidean distance between the wrist and the end effector. A sample counts as on target only when both fixed thresholds are satisfied, reducing the chance that correct-looking pointing without an adequate reach, or a reach without correct direction, will be rewarded. The proportion of on-target samples becomes a score from 0 to 100, divided into Fail below 25, Acceptable from 25 to 75 and Great at or above 75. Failed attempts must be repeated. At the end of a session, a mean score above 75 triggers promotion, while a failed-attempt ratio of 0.50 triggers demotion; demotion takes precedence if both conditions are met. The thresholds were derived in consultation with rehabilitation specialists and were not tested in patients.

Sixteen healthy adults, nine women and seven men with a mean age of 54.5 years and a standard deviation of 6.8 years, took part in two sessions. Eight entered a high-performance group and completed all repetitions as instructed. Eight entered a low-performance group in which the experimenter directed a scripted pattern of correct, failed and partial attempts during the first session. That partial pattern involved following the robot through the rightward half of the arc and resting during the excursion into the left, contralesional side. It was a controlled software test rather than natural patient behavior, and the participants were capable of correct reaching throughout. The experiment produced 120 recorded repetitions in total, with the robot’s measurements compared against BTS SMART-DX EVO stereophotogrammetric motion capture.

Across the 88 repetitions independently identified as fully correct by the motion-capture reference, measurement error increased as the arm moved farther toward the lateral workspace. The error increase was significant between Easy and the two more demanding levels, while the difference between Medium and Hard was not significant, suggesting that the largest source of difficulty arose when the arm left the frontal workspace rather than accumulating steadily with eccentricity. Median root-mean-square error for arm alignment was 3.42 degrees at Easy, 4.29 degrees at Medium and 4.39 degrees at Hard. For hand-to-end-effector distance, the corresponding medians were 1.59, 1.99 and 2.77 centimeters. Transient wrist occlusions were identified as a cause of larger errors in some trials, particularly when the wrist approached the robot’s gripper or extended laterally. Despite this, the robot and motion-capture systems assigned the same performance band in all 120 repetitions: 24 Fail, 8 Acceptable and 88 Great, with 100 percent classification accuracy and Cohen’s kappa of 1.000.

The adaptive logic also behaved as programmed for every participant. All eight members of the high-performance group were promoted from Medium to Hard after achieving no failed attempts, while all eight in the low-performance group met the demotion threshold and were moved from Medium to Easy. The high-performance group’s first-session mean score was 95.8, with individual means ranging from 86.0 to 100.0 and a fail ratio of zero. The low-performance group had a mean score of 40.7, with individual means from 39.7 to 41.8, and a fail ratio of 0.50, calculated from 24 failed attempts out of 48 first-session repetitions. After demotion, all eight low-performance participants achieved near-perfect scores in the second session, with means from 97.3 to 100.0 and no failed attempts. These results confirm that data collected asynchronously by the robot’s tracking and execution nodes reached the session aggregator and Coach node correctly. However, the performance groups were deliberately separated from the scoring boundaries: only one repetition came close to a boundary. The study therefore cannot establish how the system behaves when real patients produce gradual or fluctuating performance near the promotion and demotion thresholds. Healthy participants also do not reproduce the fatigue, impaired movement or attentional fluctuations expected after stroke.

Several important gaps remain before HemiBot could be used clinically. Its present score is motor-kinematic rather than a direct measurement of where a participant looks, so the system does not yet quantify the attentional side of neglect directly; wearable eye-tracking glasses are being explored for that purpose. Occluded samples can affect scores because the tracker does not explicitly identify every occlusion, and no minimum number of valid observations is required before an attempt is scored. The therapist’s scoreboard can override a repetition, while decisions based on session-wide statistics reduce the influence of a single anomaly, but these safeguards remain untested in a patient population. The study likewise did not establish that the robot improves neglect, that its feedback is perceived as empathetic or engaging, or that the full multi-exercise system is safe. Planned work includes usability testing, a clinical trial in people with post-stroke neglect, eye-tracking integration, gamification and more sophisticated closed-loop methods for adjusting difficulty.

Subject of Research: Technical validation of an adaptive robotic coaching system for unilateral spatial neglect rehabilitation

Article Title: HemiBot: design and technical validation of an adaptive Robotic coaching system for Unilateral Spatial Neglect rehabilitation

Article References: Soldi, R., Guerra, B. M. V., Sozzi, S., Lunghi, F., Russo, L., Schmid, M., & Ramat, S. (2026). HemiBot: design and technical validation of an adaptive Robotic coaching system for Unilateral Spatial Neglect rehabilitation. Frontiers in Robotics and AI, 13, 1950821. https://doi.org/10.3389/frobt.2026.1950821

Image Credits: AI Generated

DOI: 10.3389/frobt.2026.1950821

Keywords: HemiBot, rehabilitation robotics, unilateral spatial neglect, adaptive difficulty, mobile manipulator, ROS 2, motion capture, stroke rehabilitation, technical validation, design, technical, validation

The researchers, writing in Frontiers in Sustainable Food Systems, combined crop-specific DEA models with a pooled analysis, Tobit regression, fractional-logit robustness tests and non-parametric comparisons across the valley’s three geographical zones. Apple, pear and cherry production were analyzed separately because their underlying technologies may differ. The pooled model placed all 418 farms against a common frontier and generated the efficiency scores used in the second-stage analysis of household, institutional and regional factors. This design also allowed the team to distinguish overall technical efficiency from pure technical efficiency, which measures performance independently of production scale, and scale efficiency, which measures how closely a farm approaches the most productive scale.

DEA is a non-parametric method that constructs an empirical production frontier from observed combinations of inputs and outputs. Farms closer to that frontier are considered more efficient relative to their peers. The study used total horticultural output, measured in kilograms, as the single output. Eight inputs entered the models: landholding in marlas; urea, diammonium phosphate and potassium fertilizer in kilograms; vermicompost in kilograms; family and hired labour in man-days; plant-protection sprays in litres; and the number of horticultural assets. The input-oriented specification was chosen because growers have more discretion over the quantities and combinations of inputs they employ than over output, which is also affected by weather, biological conditions and markets.

Under the constant-returns-to-scale model, the mean score was highest for apple growers, followed by pear and cherry growers. More than half of apple farms, 51.08%, scored at least 0.80, compared with 28.07% of pear farms and 28.82% of cherry farms. Fully efficient farms—those scoring 1—accounted for 8.19% of apple growers, 6.38% of pear growers and 5.90% of cherry growers. Cherry production had the lowest average score and the widest dispersion, indicating especially uneven performance among its farms. These figures describe relative efficiency within the sample, not absolute productivity or the maximum output physically possible on each orchard.

The variable-returns-to-scale results were markedly higher: 92.05% of apple growers, 92.60% of pear growers and 92.01% of cherry growers were classified as efficient when farms were benchmarked against others of comparable scale. Scale-efficiency averages were substantially lower, at 0.8181 for apple, 0.7218 for pear and 0.6771 for cherry, and only 8.19%, 6.38% and 5.90%, respectively, reached full scale efficiency. In other words, most growers appeared to manage their available inputs effectively within their own operating size, yet many operated below or beyond the scale that would have produced the strongest frontier performance. The findings do not mean that every farm should be enlarged: scale efficiency identifies a distance from the sample’s productive scale, not the crop or size that would be optimal for an individual farm.

Farm size and production also differed sharply among crops. Apple growers reported average output of 8,488.63 kg from an average holding of 136.17 marlas, while pear growers averaged 1,643.44 kg from 36.63 marlas and cherry growers 206.15 kg from 29.27 marlas. Large standard deviations in output and input use showed substantial heterogeneity among farms. Some households cultivated more than one fruit, so the crop-specific samples included 415 apple growers, 392 pear growers and 288 cherry growers even though the household survey covered 418 farms. Sampling was multi-stage: all 10 districts were placed into North, Central and South Kashmir; two horticulture-intensive blocks were selected in each district; and commercially active households with relevant production experience were then purposively selected.

In the pooled constant-returns model, the mean farm-level score was 0.5823, with a standard deviation of 0.1623 and a range from 0.3050 to 1.0000. The lower pooled mean reflects the greater heterogeneity of a common frontier encompassing different crops, farm sizes and production systems. Tobit regression, which accommodates efficiency scores bounded between zero and one, showed that the explanatory variables were jointly significant. Family size had a positive coefficient of 0.0422, education of the household head 0.2196, artificial irrigation 0.1504 and access to institutional credit 0.0727, all with p-values below 0.01. Cold-storage availability was also positive, with a coefficient of 0.1041, but was significant only at the 10% level.

Regional location was associated with the pooled efficiency measure even after the regression accounted for household and institutional variables. Relative to Central Kashmir, coefficients were 0.1734 for North Kashmir and 0.2197 for South Kashmir, both significant at p<0.01. A fractional-logit model broadly reproduced the direction of the main relationships: family size, education, irrigation, credit and regional location remained positive, while cold storage became significant at the 5% level. Economic status was not significant in the Tobit model but negative and significant at p=0.008 in the fractional-logit specification. Market accessibility was negative in both models, significant at 10% in Tobit and 5% in fractional logit. These are conditional associations rather than evidence that any factor caused changes in efficiency.

Non-parametric tests supported the presence of geographic variation. A Kruskal–Wallis test rejected the null hypothesis that efficiency distributions were identical across the three zones at p=0.015. Pairwise Mann–Whitney comparisons with adjusted p-values found a significant difference between South and North Kashmir at adjusted p=0.018. The adjusted Central–North comparison was not significant at the conventional 5% threshold, while South–Central had an adjusted p-value of 1.000. Together with the regression results, this points to location-specific production conditions, infrastructure or services as possible explanations, although the cross-sectional design cannot isolate which underlying features produced the differences.

The study offers evidence that efficient horticulture in the Kashmir Valley may depend as much on production scale and supporting institutions as on better allocation of fertilizer, labour and other inputs. Its results associate technical efficiency with larger families, literacy among household heads, irrigation, credit, cold storage and location, while also showing that proximity to a market alone does not correspond to higher measured efficiency. Several limitations constrain generalization: the data were cross-sectional, so they do not track changes over time; they came from one valley and three fruit crops; and purposively selected commercial growers may differ from households outside the sample. The authors also note that future work could add environmental indicators, panel data and comparisons with other horticultural regions. The present results identify measurable associations and the relative importance of scale, but do not demonstrate that a specific intervention would raise output, farm income or environmental performance.

Subject of Research: Farm-level technical efficiency among apple, pear and cherry growers in Kashmir Valley

Article Title: Farm-level technical efficiency in temperate horticulture: evidence from apple, pear, and cherry growers

Article References: Sajjad, S., Dar, K. H., Ganie, A. U. H., Yousif, G. M., Salim, M., & Othman, M. S. (2026). Farm-level technical efficiency in temperate horticulture: evidence from apple, pear, and cherry growers. Frontiers in Sustainable Food Systems, 10, Article 1928386. https://doi.org/10.3389/fsufs.2026.1928386

Image Credits: AI Generated

DOI: 10.3389/fsufs.2026.1928386

Keywords: technical efficiency, Data Envelopment Analysis, Kashmir Valley, apple cultivation, pear cultivation, cherry cultivation, scale efficiency, rural credit, irrigation, sustainable horticulture, Farm-level, technical

To appreciate the significance of RT-QuIC, one must return to the conceptual problem that defined prion research for decades. Prions are infectious agents composed not of nucleic acid but of misfolded protein, a proposition famously advanced by Stanley Prusiner in 1982 and foreshadowed by theoretical work from J.S. Griffith and by radiation experiments in the 1960s suggesting that the scrapie agent could replicate without a genome. The central mechanism is templated conversion: a misfolded, aggregated form of the prion protein acts as a seed that forces the normal, protease-sensitive form of the protein to refold into the pathological conformation. This seeding model, articulated by Jarrett and Lansbury in 1993 as a kind of one-dimensional crystallization, implied that the infectious agent could in principle be detected and amplified in a cell-free system, if only the right conditions could be found.

Caughey’s laboratory delivered the crucial early proof. In 1994, working with Kocisko, Lansbury, and colleagues, his group reported the cell-free formation of protease-resistant prion protein, demonstrating that aggregates of the abnormal protein could induce the conversion of the normal protein outside any living cell. Follow-up papers in 1995 showed that these aggregates could drive the conversion of protease-sensitive prion protein to the protease-resistant state and that strain-specific properties of scrapie prions could propagate non-genetically, a finding with profound implications for how prion diversity is encoded. The same year, Caughey and Lansbury invoked the fictional ice-nine metaphor from Kurt Vonnegut’s novel Cat’s Cradle to describe the templating chemistry of scrapie infection, a seed that crystallizes everything it touches. These experiments established the mechanistic foundation on which all subsequent seed amplification technologies would be built.

The leap from proof of principle to practical detection required solving a sensitivity problem. Early cell-free conversion assays were slow and comparatively insensitive, and the field’s gold standard for detecting infectivity remained the animal bioassay, which took months or years. In 2001, Claudio Soto’s group introduced protein misfolding cyclic amplification, which used repeated rounds of sonication to amplify prion aggregates in brain homogenate. Caughey’s laboratory, together with Atarashi and colleagues, took a different route: in 2007 they described ultrasensitive detection of scrapie prion protein using seeded conversion of recombinant prion protein, and in 2010 Wilham, Orrú, and colleagues published the first full description of RT-QuIC. The assay’s elegance lay in its components. Recombinant prion protein serves as the substrate, tiny amounts of misfolded seed from a biological sample initiate its aggregation, and the aggregation is tracked in real time by a fluorescent dye, thioflavin T, that lights up as amyloid fibrils form. Gentle shaking, or quaking, accelerates the reaction in standard microplate readers.

The result was a diagnostic revolution. RT-QuIC could detect prion seeding activity with sensitivity comparable to animal bioassays but in hours rather than months, and it required no infectious material beyond the sample itself. Successive refinements pushed the technology further. In 2014, Orrú and colleagues reported in the New England Journal of Medicine that nasal brushings could be used to diagnose Creutzfeldt-Jakob disease with remarkable accuracy, and in 2015 the same team showed that cerebrospinal fluid testing with RT-QuIC achieved high sensitivity and specificity for sporadic CJD. International ring trials confirmed the assay’s robustness across laboratories, and second-generation protocols standardized the substrate and reaction conditions. Today, RT-QuIC on cerebrospinal fluid is incorporated into diagnostic criteria and surveillance guidelines for human prion disease issued by bodies such as the CDC and the European Centre for Disease Prevention and Control, and it has been extended to an astonishing range of easily accessible tissues, including olfactory mucosa, skin punch biopsies, hair roots, tear fluid, and even placental tissue.

Perhaps the most consequential extension of the platform came when Caughey’s group and others realized that the same seeding logic applies to other misfolded proteins implicated in neurodegenerative disease. Alpha-synuclein, the protein that aggregates in Parkinson’s disease, dementia with Lewy bodies, and multiple system atrophy, was adapted to RT-QuIC formats beginning around 2013, with parallel development of the related alpha-synuclein seed amplification assay by Shahnawaz and colleagues. Groveman and colleagues published a rapid, ultrasensitive alphaSyn RT-QuIC in 2018 for quantifying pathological alpha-synuclein seeds in brain and cerebrospinal fluid. The clinical payoff has been dramatic: large multicenter studies, including analyses within the Parkinson’s Progression Markers Initiative, demonstrated that cerebrospinal fluid alpha-synuclein seed amplification can identify Parkinson’s disease with high accuracy, and longitudinal work published in The Lancet Neurology in 2025 showed that the kinetic measures of the assay carry diagnostic and prognostic value. Researchers have since detected alpha-synuclein seeds in skin, olfactory mucosa, duodenal biopsies, and serum, opening the door to minimally invasive diagnosis at early stages of disease.

The same logic has been applied to tau, the microtubule-associated protein whose aggregates define Alzheimer’s disease and a family of frontotemporal dementias collectively called tauopathies. Kraus and colleagues showed in 2019 that RT-QuIC could selectively detect tau aggregate conformers of Alzheimer’s disease, and Metrick and colleagues developed a single ultrasensitive assay capable of discriminating tau aggregates of Alzheimer’s disease from those of Pick disease. Salt-modulated amplification protocols now allow classification of tauopathies directly from brain homogenates, and tau seeding activity has been detected in skin biopsies, where it helps differentiate tauopathies from synucleinopathies. The platform has expanded further still: TDP-43 seeds have been measured in cerebrospinal fluid and olfactory mucosa of patients with amyotrophic lateral sclerosis and frontotemporal dementia, superoxide dismutase 1 aggregates have been detected in neural tissue and cerebrospinal fluid from ALS cases, and even misfolded insulin has been targeted with a dedicated RT-QuIC assay.

Beyond human diagnostics, RT-QuIC has become an indispensable tool for surveillance and control of animal prion diseases, particularly chronic wasting disease in deer and elk, which continues to spread across North America and beyond. The assay enabled antemortem detection of prions in saliva, urine, blood, and feces of infected cervids, longitudinal studies of prion shedding, and detection of prions on environmentally relevant surfaces and in venison processing environments. Ear-notch testing of free-ranging and farmed deer is now feasible, and recent work has found prions in the blood of healthy-appearing white-tailed deer, underscoring the challenge of silent carriage. The assay has also been turned on itself in a constructive way: because it quantifies residual seeding activity, it has become a standard method for evaluating prion decontamination procedures, testing disinfectants such as hypochlorous acid, sodium hypochlorite, and commercial formulations against multiple prion strains on laboratory surfaces.

The technology continues to evolve in directions that Caughey championed: higher throughput, greater sensitivity, and broader accessibility. Hofmeister ion comparisons achieved million-fold sensitivity enhancements in seed amplification assays for biospecimens, magnetic particle extraction concentrates prions from large-volume samples, and nanoparticle-enhanced Nano-QuIC has overcome inhibitors in blood to detect alpha-synuclein seeding activity in Parkinson’s patients. Microfluidic Micro-QuIC platforms promise rapid on-site amplification and visual detection, while digital seed amplification assays bring single-molecule counting to TDP-43 quantification in cerebrospinal fluid. In parallel, the assay has become a workhorse of therapeutic research, serving as a screening tool for anti-prion drug candidates and as a readout in human cerebral organoid models of Creutzfeldt-Jakob disease, where recent work has identified oligosaccharyltransferase complex inhibition as a promising treatment strategy for rodent and human prions.

What emerges from the review is a portrait of a scientific legacy built on collaboration, standardization, and an almost obsessive attention to the practical details of assay performance, from substrate selection, such as the remarkably versatile bank vole prion protein, to buffer chemistry and shaking parameters. The authors are careful to note that seed amplification technology has been developed through the seminal contributions of many eminent scientists and that the breadth of the field now extends far beyond any single laboratory or any single review. Yet the through-line is unmistakable. A question that began with the strange chemistry of scrapie, pursued for decades by Byron Caughey and his collaborators, has grown into a family of ultrasensitive assays that now detect, discriminate, and quantify the pathological protein seeds of prion disease, parkinsonism, tauopathy, and motor neuron disease. As the field moves toward standardization, data sharing, and clinical deployment of alpha-synuclein and tau seed amplification assays, the tree of discoveries that grew from a single prion seed continues to spread its branches across medicine.

Subject of Research: Development and legacy of the RT-QuIC seed amplification assay for ultrasensitive detection of prions and other misfolded proteins

Article Title: From a prion seed to a tree of discoveries: Byron Caughey’s RT-QuIC legacy in protein misfolding research

Article References: Groveman, B. R., Vascellari, S., Bongianni, M., Hughson, A. G., Zanusso, G., & Orrú, C. D. (2026). From a prion seed to a tree of discoveries: Byron Caughey’s RT-QuIC legacy in protein misfolding research. Acta Neuropathologica, 152(1), Article 44. https://doi.org/10.1007/s00401-026-03094-3

Image Credits: AI Generated

DOI: 10.1007/s00401-026-03094-3

Keywords: RT-QuIC, prion disease, Byron Caughey, protein misfolding, alpha-synuclein, tau, Creutzfeldt-Jakob disease, Parkinson's disease, seed amplification assay, chronic wasting disease, TDP-43, biomarkers

The study, conducted by Fei Song and Michael Dunn of the Centre for Biomedical Ethics at the National University of Singapore together with Zhenni Hu of Nanjing Normal University, examined 43 articles indexed in the Chinese Social Sciences Citation Index along with four policy reports. The analysis reveals that Chinese scholars have made substantial contributions to policy and governance discussions around trustworthy AI, but that these contributions remain conceptually scattered, with different disciplines approaching the question of trust from radically different angles. The researchers set out to synthesize this dispersed knowledge into a coherent conceptual account, and in doing so they arrived at a framework that treats trustworthiness not as a property of algorithms alone but as an emergent condition spanning technology, institutions, and human relationships.

The first layer of the framework concerns technical foundations. This is the layer most familiar to engineers and computer scientists, encompassing properties such as reliability, robustness, transparency, explainability, fairness, and safety. Chinese-language scholarship in this domain is extensive and spans a striking range of applications. Researchers have examined explainable learner models as the technical key to trustworthy personalized learning, explored blockchain-driven approaches to financial security intelligence, and investigated how algorithmic transparency in personalized recommendation systems shapes users’ perceptions of trustworthiness. Work on data annotation governance has highlighted backstage risks in the labor-intensive processes that feed machine learning systems, while archival science perspectives have been brought to bear on the question of algorithm provenance, asking how the origins and histories of algorithms can be documented and verified. These contributions demonstrate that the technical layer of trustworthiness is not merely a matter of model architecture but extends deep into the data pipelines, documentation practices, and supply chains that underpin AI systems.

The second layer addresses governance and regulatory arrangements. Here the Chinese literature is particularly distinctive, reflecting the country’s active policy discourse on AI regulation. Scholars have assessed pathways for trustworthy AI legislation, explored innovation-friendly regulatory designs that balance safety with technological development, and analyzed governance practices in terms of models, actors, objects, and tools. Policy documents play an important role in this layer, including the Ministry of Science and Technology’s 2019 governance principles for developing responsible AI, a white paper on trustworthy artificial intelligence produced by the China Academy of Information and Communications Technology, and more recent industry ecosystem reports and governance surveys. The review shows that Chinese scholars have been grappling with the same fundamental questions that animate regulators in Europe and the United States: how to translate abstract ethical principles into enforceable rules, how to assign liability when AI systems fail, and how to design oversight mechanisms that keep pace with rapidly evolving technology. One striking example is legal scholarship on liability when human–machine co-piloting fails, examining the negligence of safety operators through the lens of the principle of reliance.

The third layer, and the one the authors argue is most often neglected in purely technical accounts, concerns the conditions of human–AI interaction under which trust is formed and calibrated. The Chinese-language literature here is remarkably rich, drawing on empirical psychology, communication studies, and philosophy. Researchers have studied trust in automated vehicles, safety trust in intelligent domestic robots, and human–AI mutual trust in the era of artificial general intelligence. A dual-pathway model of trust calibration distinguishes between trust dampening and trust promoting mechanisms, offering a nuanced picture of how users adjust their reliance on machines. Other studies examine how anthropomorphism affects trust in AI agents, with a meta-analysis synthesizing evidence on how humanlike features in AI systems influence trust and how contextual factors moderate that effect. Work on large language models has explored how machine hallucinations impair human–machine trust and what recalibration mechanisms might restore it, a question of obvious urgency as chatbots and AI assistants become ubiquitous.

What emerges from this synthesis is the insight that these three layers are not independent. An AI system can be technically excellent, with high accuracy and robust performance, and still fail to be trustworthy if the governance arrangements around it are inadequate or if the conditions of human interaction systematically miscalibrate trust. Conversely, strong institutions and well-designed interaction conditions cannot compensate for fundamentally unreliable technology. The authors’ central claim is that an AI system is trustworthy only if it meets the requirements at all three layers simultaneously. This conjunctive structure has practical implications: it means that trustworthiness cannot be certified by any single test, audit, or label, but requires integrated assessment across the full stack from model internals to regulatory environments to the design of the encounters between humans and machines.

The framework also speaks to a long-running philosophical debate about whether trust in AI is even coherent. Some Western philosophers argue that trust, properly understood, requires mutual expectations and moral agency that machines cannot possess, and that what we call trust in AI is really a form of reliance or confidence. Others contend that it is entirely possible to trust medical AI systems and that denying this misunderstands the functional role of trust in clinical decision-making. The Chinese literature adds distinctive voices to this debate, with scholars asking directly whether artificial intelligence can serve as a trustee and defending the notion of trustworthy AI from a logical perspective. The review also engages with a network account of trustworthy AI, which holds that trustworthiness is distributed across the networks of actors, artifacts, and institutions surrounding an AI system rather than residing in the system alone. The three-layer framework can be read as a structured elaboration of this network idea, specifying the technical, institutional, and interactional nodes that such a network must contain.

The methodological significance of the study lies partly in its source base. Scholarship published in Chinese remains largely absent from mainstream academic debates, a gap reinforced by publication incentives that push Chinese humanities and social science researchers toward international English-language venues. By drawing on the Chinese Social Sciences Citation Index, the authors accessed a body of work that includes empirical studies using artificial neural networks and Q methodology to map the dimensions of human–AI trust, philosophical analyses grounded in Confucian traditions, and domain-specific investigations in education, journalism, healthcare, and finance. The result is a more globally representative picture of what the research community actually knows about trustworthy AI. The authors also note distinctive features of the Chinese context, including a practice-oriented view of science and technology that shapes how trustworthiness is framed, and a governance culture in which state-led principles, industry white papers, and academic analysis interact closely.

The study concludes by identifying three directions for future research. First, the conceptual fragmentation of the field calls for continued theoretical work to integrate technical, governance, and interactional perspectives into unified accounts of trustworthiness. Second, the rapid emergence of generative AI and large language models creates new empirical challenges, from hallucination-induced trust erosion to the trust dynamics of human–machine dialogue, that demand updated frameworks. Third, cross-linguistic and cross-cultural comparative research is needed to test whether the three-layer structure holds across different scholarly traditions and regulatory cultures, or whether trustworthy AI means genuinely different things in different contexts. As AI systems increasingly mediate news consumption, medical diagnosis, education, and financial services, the stakes of getting trustworthiness right could hardly be higher. This review makes the case that answering the question well requires listening to conversations that have been happening in Chinese all along, and that the resulting framework, grounded in arguments advanced in the Chinese context, offers a more robust conceptual account of how disparate claims about trust fit together than either technical or governance approaches alone can provide.

Subject of Research: Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship

Article Title: Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship

Article References: Song, F., Hu, Z., & Dunn, M. (2026). Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship. AI &amp; SOCIETY. https://doi.org/10.1007/s00146-026-03366-2

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03366-2

Keywords: Trustworthy, beyond, technical, three-layer, framework, informed, Chinese-language, scholarship, scientific research