The research, conducted by Roos Scheffers, Floris Bex, and Matthieu Brinkhuis of Utrecht University, set out to test whether explanation definitions drawn from the computational argumentation literature align with real human explanation behaviour. Computational argumentation represents knowledge as sets of arguments connected by attack relations: one argument can undercut another, and a third argument can defend the first by attacking its attacker. This structure mirrors the argumentative character of human reasoning, in which people naturally seek support for conclusions, weigh objections, and mentally prepare rebuttals in advance of a challenge. Because of this cognitive grounding, the field has long assumed that argumentation-based explanations would feel natural to users. But, as the authors point out, that assumption has been largely untested. Prior empirical work had examined how people evaluate arguments and attack relations, yet only a single earlier study had looked at how people actually explain arguments.

At the heart of the study are three formally defined types of explanation. A sufficient explanation contains the set of arguments needed, together with the topic argument, to guarantee its acceptance against all attackers. A compact explanation is a sufficient one with no redundant members, meaning no proper subset of it would still do the explanatory work. A minimal explanation is the smallest sufficient explanation measured purely by the number of arguments it contains. These categories are nested: every minimal explanation is compact, and every compact explanation is sufficient. Each type embodies a different degree of selectivity, the cognitive-science-inspired principle that good explanations should not overwhelm the recipient with information but should instead zero in on what matters for the conclusion being explained.

To find out which of these definitions best captures human intuition, the researchers recruited 301 participants through the online crowdsourcing platform Prolific, drawing English-fluent adults from 42 countries. The experiment was preregistered, its sample size determined by a power analysis, and it was approved by the Utrecht University Science-Geo Ethics Review Board. Participants worked through eight argumentative scenarios drawn from a pool of twenty that spanned domains including criminal investigations, environmental policy, peer review, and school projects. Each scenario presented a topic argument assumed to be true, one or more counterarguments attacking it, a set of defending arguments that attacked the counterarguments, and one deliberately unrelated argument sharing the same context but playing no role in the dispute.

The task itself was elegant in its simplicity. Participants were asked to explain the conclusion of the topic argument, given the counterarguments, by ticking the boxes of the arguments they believed explained it. They could select as many or as few arguments as they wished, provided they chose at least one. Their selections were then compared against the three formal explanation types and against two statistical baselines: a naive baseline assuming every possible explanation is equally likely to be picked, and a simulated baseline weighted by the observed distribution of explanation lengths among participants. The preregistered hypotheses predicted, in increasing order of selectivity, that people would prefer sufficient, then compact, then minimal explanations more often than chance would suggest.

The results confirmed the hypotheses, though with revealing nuances. Across both argumentation frameworks used in the study, one smaller and one larger, explanations fitting the sufficient, compact, and minimal types were chosen significantly more frequently than the baselines predicted. In the smaller framework, where the minimal explanation consisted of a single argument, participants embraced all three types enthusiastically. In the larger framework, which added a third counterargument and required at least two arguments for a complete defence, the picture shifted: minimal and compact explanations still beat both baselines, but sufficient explanations only outperformed the baseline that accounted for participants’ strong preference for brevity. The take-home message, the authors conclude, is that people prefer short explanations built from arguments directly related to the topic.

That preference for brevity proved remarkably stubborn. In the smaller framework, participants’ explanations averaged just 1.48 arguments, with 62 percent consisting of a single argument. In the larger framework, even though more information was available and a complete defence demanded more arguments, the average grew only to 1.74, an increase of roughly 25 percent. More than half of participants still chose a single-argument explanation. Critically, when they did choose lone arguments, they almost never chose the unrelated one; the distractor argument was picked only about 4 to 7 percent of the time, far below the rate of any relevant argument. This tells the researchers that participants were not simply being lazy. They could tell which arguments mattered and deliberately excluded the ones that did not.

The most striking finding concerned partial explanations. In the larger framework, the majority of participant responses fit none of the three formal types, and most of these consisted of a single defending argument that repelled only one of the topic argument’s several attackers. Participants appeared to identify the argument they judged most important, often the one that fended off multiple attackers or neutralised a unique threat, and stopped there, even though formally complete justification required more. The authors interpret this pattern through the lens of cognitive load theory, the well-established idea that human working memory has limited processing capacity. As the argumentation scenario grew more complex, participants did not scale up their explanations proportionally; instead they simplified, offering shorter and more selective answers rather than absorbing the extra cognitive cost of a full defence. This echoes earlier findings that people adopt simpler reasoning strategies when formal argumentation frameworks become more complicated.

The implications for explainable AI are significant. On one hand, the study validates the field’s foundational intuition: argumentation-based explanation definitions, at least those centred on sufficiency, compactness, and minimality, do capture genuine regularities in human explanatory behaviour, particularly the drive toward relevant, related arguments. On the other hand, the results expose a gap. Formal definitions that require complete, admissible explanations produce output that is systematically longer than what people naturally offer. To close that gap, the authors argue, future work should develop explanation definitions that permit selective, even formally incomplete explanations while preserving as much formal rigour as possible. Such definitions would yield AI explanations that feel less like exhaustive legal briefs and more like the crisp, pointed answers humans actually give.

The study also maps out its own limits. Individual participants varied widely, with about two-thirds changing their explanation length across scenarios while a consistent third always picked a single argument. Responses differed measurably across scenarios, hinting that context and wording shape how people explain, even though no single scenario deviated significantly from the overall pattern in follow-up tests. The experimental setting featured low stakes and no time pressure, and the participants were laypeople rather than domain experts; preferences might shift in high-pressure professional environments such as courtrooms or medical settings, where explanations of AI systems often matter most. The researchers have released their twenty scenarios and both argumentation frameworks as open materials, hoping they will serve as benchmarks for testing new explanation definitions against the explanation behaviour of real human reasoners, a step they see as essential for building AI that is both formally sound and psychologically realistic.

Subject of Research: Empirical testing of human explanation preferences against formal explanation definitions in computational argumentation

Article Title: Empirically Testing Explanation Preferences in Computational Argumentation

Article References: Scheffers, R., Bex, F., & Brinkhuis, M. (2026). Empirically Testing Explanation Preferences in Computational Argumentation. Cognitive Computation, 18(1), Article 109. https://doi.org/10.1007/s12559-026-10654-y

Image Credits: AI Generated

DOI: 10.1007/s12559-026-10654-y

Keywords: computational argumentation, explainable AI, human reasoning, explanation selection, cognitive science, argumentation frameworks, empirical study, cognitive load, selective explanations, artificial intelligence, XAI, formal logic

The work, published open access in the Journal of Ambient Intelligence and Humanized Computing, responds to a long-standing problem in environmental communication. Conventional charts are excellent for analysis, but research on psychological distance shows that large-scale, slow-moving phenomena such as climate change often fail to register as personally urgent when they are presented as abstract numbers. Eco-visualization, a practice first articulated in museum contexts and later explored in energy-feedback research, attempts to close that gap by blending data representation with aesthetic, narrative, and experiential elements. The catch has always been skill: designing these hybrid images demands both creative sensibility and technical competence in data processing, segmentation, and color mapping, a combination that excludes many talented designers.

The team’s answer is a structured, six-step Human-AI co-creative pipeline. The process begins with generative image creation: the user describes a dataset topic, and a text-to-image model proposes visually meaningful scenarios, such as a flooded coastal city for sea-level data or a bleached coral reef for ocean acidification. The user can accept a suggestion, request variations, or start over, iteratively steering the AI toward a contextually appropriate scene. Next comes dataset configuration, in which the user uploads structured quantitative data in CSV or TSV format; the system parses the file, displays columns as interactive chips, and pre-computes minimum and maximum values that will later define the color scale.

The third step is where the machine-learning engineering becomes visible. The system integrates Meta AI’s Segment Anything Model, or SAM, to automatically analyze the generated image and produce pixel-level masks for each identifiable object or region. Three Vision Transformer backbones of increasing size, ViT-B, ViT-L, and ViT-H, are available, but the interface deliberately hides that choice: ViT-B runs by default, and larger variants sit behind an optional advanced control. If the automatic masks miss something, a lasso tool lets users draw custom regions, and a mask editor supports merging segments through Boolean union operations, undoing recent changes, or resetting to the original output. The design philosophy is that users should evaluate the masks that matter to their visualization, not choose model architectures.

Color is treated with scientific seriousness. The system encourages palettes validated for perceptual uniformity and accessibility, such as the scientific color maps developed by geoscientist Fabio Crameri and colleagues, whose 2020 Nature Communications paper documented how arbitrary gradients can mislead viewers and distort quantitative interpretation. Users can also build custom palettes from start and end colors, and the AI suggests alternatives derived from the dominant colors of the generated image so the visualization blends naturally into its scene. During overlay generation, the chosen palette is mapped linearly onto the dataset’s value range, and the gradient is applied only inside the masked region using alpha blending, with independent hue, saturation, and lightness coefficients giving designers fine-grained control over how strongly data colors replace the original pixels. A separate edge-smoothing parameter softens boundaries between modified and untouched areas.

To understand whether this pipeline actually works for its intended audience, the researchers ran a remote user study with 13 participants, twelve of them from design backgrounds, chosen precisely because they combine visual expertise with limited experience in data-driven toolchains. Participants received no training, only a brief written introduction and a guided task: generate a mountain landscape, load monthly atmospheric CO2 data, merge masks if desired, colorize only the lake region with a scientific palette, and export the result. The evaluation combined the NASA Task Load Index for cognitive load, the short User Experience Questionnaire for pragmatic and hedonic quality, open-ended feedback, and a demographic section, all collected anonymously in compliance with GDPR.

The results were strikingly divided. Cognitive load, measured on a 0 to 100 scale, averaged 59 with a standard deviation of 16.1, squarely in the high range, and the spread suggests the mental burden varied considerably from person to person. Pragmatic quality, the UEQ-S dimension capturing usability and task effectiveness, scored a mean of minus 0.96, below the minus 0.8 threshold that marks a problematic rating. Items probing whether the system felt easy and clear scored poorly, and participants reported that the browser froze during segmentation, a consequence of running the model locally on CPU. One participant described the HSL adjustments as monotonous and struggled to grasp what the parameters actually changed, hinting that the interface failed to communicate the rationale behind its controls.

Yet hedonic quality told the opposite story, averaging 1.33, a clearly positive evaluation. Participants rated the system as highly inventive and on the leading edge, with scores of 6.23 and 5.69 respectively on those items, indicating that the core concept of embedding environmental data inside emotionally resonant imagery resonated strongly even as the interaction stumbled. The overall UEQ-S score of 0.18 landed in the neutral range, with the system’s conceptual appeal only partially compensating for its usability problems. The authors argue this gap exposes a fundamental tension in Human-AI co-creation: granting users expressive control increases creative freedom but inflates cognitive demand when the purpose of each parameter is not immediately obvious, pointing toward design strategies like progressive disclosure and contextual explanation.

The team is candid about the study’s limits. The sample was small, deliberately composed of designers, and constrained to a prescribed task with a shared dataset, so the findings speak to first-use learnability rather than open-ended creative practice. CPU-based segmentation introduced latency that degraded perceived responsiveness, and the AI’s role, while framed as collaborative, remains fundamentally reactive: it responds to prompts but does not learn from users across sessions or proactively shape the creative process, placing the current system closer to AI-assisted design than to fully bidirectional co-creation. Future work will move segmentation to GPU-accelerated or cloud inference, add contextual guidance for parameter controls, and run ecologically valid studies in which participants choose their own datasets, audiences, and visual narratives, ideally involving environmental scientists, educators, and communication professionals alongside designers.

Even with those caveats, the contribution is concrete and timely. At a moment when generative models are flooding the internet with images, this research asks a sharper question: can those models help make invisible environmental processes visible without sacrificing scientific integrity? By pairing perceptually reliable color science with iterative human steering, the pipeline sketches a workflow in which data and imagination reinforce rather than compete with each other. The mountain lake that darkens as carbon dioxide climbs is not a replacement for the chart behind it, but it is a complement that speaks to a different part of the mind, and the study’s mixed verdict, captivating concept, demanding interface, maps the road that co-creative tools must travel before that vision becomes routine practice in climate communication.

Subject of Research: A human-AI co-creative pipeline for designing data-driven eco-visualizations

Article Title: Co-creating eco-visualizations: a human–AI collaborative pipeline for data-driven visual design

Article References: Ceccarini, C., Bottari, B., Liçaj, A., Matteucci, E., & Delnevo, G. (2026). Co-creating eco-visualizations: a human–AI collaborative pipeline for data-driven visual design. Journal of Ambient Intelligence and Humanized Computing. https://doi.org/10.1007/s12652-026-05128-w

Image Credits: AI Generated

DOI: 10.1007/s12652-026-05128-w

Keywords: eco-visualization, human-AI co-creation, data visualization, generative AI, image segmentation, Segment Anything Model, environmental communication, cognitive load, user experience, scientific color palettes, climate data, co-creative systems