Every Brain Has a Cognitive Pattern

If people do not perceive and represent the world in the same way, why should we expect a single interface to feel equally natural and understandable to everyone?

This article begins with a simple observation: people differ in how they receive, represent and organise information and solve problems. Research in cognitive psychology, neuroscience and design shows that individual differences in cognitive styles, spatial abilities, visual and verbal representation, and problem-solving strategies are real and measurable. At the same time, current evidence does not allow us to claim that a complete, universal and immutable “cognitive fingerprint” has been demonstrated for every person. Accordingly, this article presents the “uniqueness of cognitive patterns” as a testable hypothesis rather than an established fact.

The article’s central argument is that a substantial part of what everyday language calls “taste” may be more than a freely chosen, superficial preference. It may instead emerge from interactions among sensory perception, attention, memory, language, culture, experience, expertise, mental schemas and biological predispositions. Examples such as colour vision deficiency, linguistic differences in the categorisation of blue in Russian, research on visual-spatial memory in Australian Aboriginal children, and evidence for the heritability of some spatial abilities suggest that the space of choice does not necessarily begin from the same point for everyone.

On this basis, the proposed concept of “interface cognitive compatibility” (Cognitive Compatibility) is introduced as an operational framework for this article: the degree to which the presentation of a product or interface aligns with the way a user processes information more quickly and with less friction. This idea does not arise in a vacuum. Cognitive Fit theory, adaptive interface research, Website Morphing, automatic identification of cognitive style from digital interactions, and CAPTCHA studies have each previously tested parts of this path [19-24]. This article proposes integrating these research strands with today’s capabilities in artificial intelligence and Generative UI: a few very brief, indirect choices, including free selection among visual, spatial and rule-based Micro-Puzzles, produce a cognitive signal. The user’s preference between two different representations of the same functionality is then measured, and the relationship between the two is learned at population scale. The aim is not to assign definitive labels to people, but to build a statistical space linking choice patterns and interface patterns, so that in the future AI can adapt the presentation of the same functionality to how the user understands it, alongside needs-based personalisation.

Keywords: cognitive style, cognitive fingerprint, cognitive compatibility, adaptive user interface, artificial intelligence, Object-Spatial-Verbal, Cognitive CAPTCHA, interaction design

Level Meaning Example in this article

Strong support Individual cognitive differences, the existence of Object/Spatial/Verbal dimensions, the relative stability of some abilities, and the effects of context and experience OSIVQ research, cognitive reviews and the meta-analysis of ability stability

Limited/context-dependent support Effects of language on some perceptual tasks and cultural or environmental differences in cognitive strategies Russian blues and Kearins’s studies

Article hypothesis A comprehensive Cognitive Fingerprint for each individual and the possibility of predicting interface compatibility from a few short Micro-Puzzles Cognitive Compatibility and Cognitive CAPTCHA

When people encounter a product, design, artwork or even a scientific theory, differences of opinion are taken for granted. One person finds an interface “clear and beautiful”; another calls the same interface “superficial and ambiguous.” One enjoys diagrams and structural maps, while another does not feel they understand the subject without seeing the final result. We usually group all these differences under a general term: taste. Yet rather than explaining the cause, this word often simply names the observed outcome.

Research on cognitive styles shows that individuals differ in how they acquire and process information. These differences can be understood as patterns of cognitive adaptation to the environment, formed on the basis of biological predispositions and modified by environmental demands, experience and culture [1]. The starting point of choice is therefore not identical for everyone. Before someone can “like” something, they must see it, distinguish it, attend to it, connect it to prior knowledge and construct a mental model of it.

The article’s central hypothesis: what we call “preference” is sometimes the output of a deeper chain of perception, representation, experience and cognition, rather than simply a free choice among equally accessible, equivalent options.

This article does not seek to place people in a few fixed boxes. Instead, it describes a multidimensional space in which each individual may have a different combination of cognitive preferences and abilities. Nor is the ultimate aim personality diagnosis; it is to examine whether the relationship between “how we understand” and “how a product is presented” can be measured and applied in intelligent design.

Differences in mental representation can first be seen in the creative process. When several people are asked to imagine a new product, one may first see the final product’s appearance; another may simulate its movement mechanism and the connections between its parts; a third may begin with the production route, equipment and assembly; and a fourth may formulate the problem in terms of goals, constraints and verbal rules. This is not simply a difference in occupation, although expertise and experience reinforce it.

The three-dimensional Object-Spatial-Verbal model and the OSIVQ questionnaire rest on a distinction: “object imagery” concerns visual detail, colour, form and image quality; “spatial imagery” focuses more on relationships, transformations, movement, orientation and structure; and verbal processing relies on words, propositions and linguistic reasoning [2]. Earlier research has also shown that being a visualizer is not a single dimension: iconic/object imagery can be distinguished from schematic/spatial imagery [3].

In design science, the Function-Behaviour-Structure framework similarly shows that a designed object can be represented in terms of “what it was made for,” “how it behaves” and “what structure and relationships constitute it” [12]. This framework is not itself a classification of people, but it shows that even a single product can be viewed at different cognitive levels.

Need / Function Behaviour and mechanism / Behaviour Structure / Structure Final form / Representation Production process / Process

Figure 1 - Possible entry points for the mind into a design problem; these paths are not necessarily linear or mutually exclusive.

Design research also shows that professional designers continually move between problem and solution, with the definition of the problem evolving as the solution develops [13]. Conversely, showing an example early can create Design Fixation, constraining the designer to the initial example’s features [14]. The mind’s entry point into a problem can therefore alter not only taste, but also the space of imaginable solutions.

This article does not claim that contemporary science has proved that every person possesses a comprehensive, wholly unique and immutable cognitive profile analogous to a skin fingerprint. Such a claim exceeds the available evidence. Several lines of research do, however, provide good reasons to take the hypothesis seriously and regard it as testable.

At the brain level, Finn and colleagues showed that patterns of functional brain connectivity can be sufficiently individual-specific to identify people in fMRI data; they used the term functional connectome fingerprint [4]. This finding does not directly establish a “cognitive fingerprint,” because a measure that identifies someone from brain connections is not necessarily the same measure that predicts their preferences or behaviour.

At the level of cognitive behaviour, Schulz and colleagues showed in a pseudo-random number generation task that individual patterns of choice and pattern inhibition could distinguish “the same person” from “another person” with an AUC of around 96.5 percent based on 300 digits. The pattern remained stable over one week; the authors called it a cognitive fingerprint [5]. This is a very interesting result, but it concerns a specific task and a limited sample and should not be generalised to the whole of personality or cognition.

Meanwhile, a meta-analysis of 205 longitudinal studies involving 87,408 participants shows that cognitive abilities have high rank-order stability from late adolescence into adulthood. For a 20-year-old over a five-year interval, average stability was reported at around ρ=.76 [6]. “Relatively stable” is therefore more accurate than “fixed.”

Proposed scientific formulation: each individual may have a multidimensional, relatively stable profile of cognitive preferences, abilities and strategies. The combination of these dimensions may be individual-specific to some degree and enable an operational Cognitive Fingerprint to be defined. This is a research hypothesis, not a proven assertion.

A perceptual example can clarify the limits of the word taste. A person with colour vision deficiency does not usually lose all colour perception. In the most common forms, distinguishing certain colours becomes more difficult; in rarer forms, the range of colour vision may become very limited [18]. If such a person is indifferent to part of the colour spectrum when choosing a shirt, or consistently favours a group of colours, their perceptual system has constrained some of their behaviour before it reaches “preference.”

This example does not equate cognitive differences with sensory impairment. It merely illustrates a principle: the range of options from which the mind can actually choose is not necessarily the same for everyone. The final choice emerges after information passes through multiple filters.

Sensory perception Attention Categorisation Memory and experience Mental model Valuation Choice

Figure 2 - A simplified chain showing that “preference” emerges at the end of several processing stages.

Psychological schemas, culture, language, professional experience and the immediate goal must also be added to this chain. Thus, when a user says “I don’t like this interface,” the deeper meaning is not always “I don’t like this colour.” Sometimes the less obvious meaning may be “this interface presents information in a way that makes it harder for me to turn it into a mental model.”

Language does more than attach names to experience; in some circumstances, it can also change how experience is categorised and how quickly it is processed. A well-known example is the Russian vocabulary for blue. Russian makes a basic, obligatory lexical distinction between light blue (goluboy) and dark blue (siniy), whereas English groups both within the general category blue.

In 2007, Winawer and colleagues showed that Russian speakers performing a rapid colour-discrimination task distinguished two shades of blue faster when they belonged to different linguistic categories, siniy/goluboy, than when both belonged to the same category. English speakers showed no such categorical advantage with the same stimuli. More importantly, a verbal interference task eliminated this advantage, whereas spatial interference did not [7].

This result should not be exaggerated into “Russians see colours that others do not.” A replication study by Martinovic and colleagues in 2020 did not reproduce the same speed advantage across the siniy/goluboy boundary and showed that the effect is sensitive to context, stimulus distribution and experimental conditions [8]. The conservative conclusion is therefore that language can affect categorisation and perceptual processing under some conditions, but the effect is neither absolute nor immutable.

For this article’s argument, that is sufficient: what someone finds “obvious” or “different” at the moment of choice is not solely a product of the external object. Their conceptual and linguistic system may also be part of the process.

Genetics requires greater care. Judith Kearins’s 1981 research showed that a group of Aboriginal children from Australian desert regions performed better than White Australian children on visual-spatial memory tasks, including recalling objects’ locations. Their behavioural patterns also differed: Aboriginal children used visual strategies more often, while many White children relied on verbal naming strategies [9].

The initial interpretation proposed an “environmental pressures” hypothesis: spatial skills may be more valuable in an environment where navigation and remembering spatial landmarks are particularly important. However, the same line of research did not support the simple account that treats this acquired ability as directly genetic and transmitted to children. In 1986, Kearins explicitly noted that the initial genetic assumption had rested on the mistaken belief that non-traditional Aboriginal children had been raised like White children, whereas many traditional child-rearing practices and environmental experiences had been retained [10].

Behavioural genetics, on the other hand, shows that some spatial abilities do have substantial heritability. A meta-analysis of twin studies reported the average genetic component of spatial ability at around a²=.61 [11]. This does not mean that 61 percent of an individual’s spatial ability “came from their genes”; heritability is a population-level measure of the origins of differences within a particular environment.

The appropriate model for this article is not “genetics or environment,” but their interaction: biological predisposition × neural development × experience × upbringing × culture × language × expertise.

The idea of a “cognitive pattern” must therefore be dynamic. If cognitive differences have both biological and environmental roots, an adaptive system should not permanently confine a user to a fixed category either; it should revise its model in response to new behaviour.

Suppose a company manager requests a website during a meeting. A designer, programmer, product manager and marketing lead may look at the same website-builder panel, yet each find different elements salient: form and aesthetics, technology and architecture, process and control, or business outcomes. Some of this difference comes from their roles and interests, but some may also relate to how information is represented.

First impressions of interfaces also form very quickly. Lindgaard and colleagues showed that ratings of web pages’ visual appeal after a 50-millisecond exposure correlated with ratings after longer exposure [15]. This finding concerns visual appeal and should not be generalised to a complete understanding of usability, but it shows that the initial encounter can take shape before in-depth analysis.

Under these conditions, an interface centred on a final preview, large images and direct manipulation may quickly appear “understandable” to someone inclined towards Object Representation. The same interface may seem like a Black Box to someone who needs to see structure, dependencies and the operating path before trusting the system. The difference in judgement here is not necessarily disagreement about the actual quality of the functionality; it may concern the cost of converting the external display into a mental model.

The article’s conceptual proposal: instead of asking “Is this interface good or bad?”, we can ask “For which ways of processing information does this interface create more or less cognitive friction?”

In this article, Cognitive Compatibility, or “interface cognitive compatibility,” describes the degree of alignment between how an interface presents information and how the user processes it. This concept should be regarded as distinct from, but related to, the established theory of Cognitive Fit. In 1991, Vessey showed that even when a graph and a table contain identical information, their representation can foreground different kinds of processing, spatial or symbolic. When the representation matches the task’s cognitive demands, problem-solving becomes faster and more accurate [19]. Cognitive Compatibility takes a hypothetical step further here, asking whether “representation fit” relates only to the type of task or may also relate to more stable differences between users. This usage is not yet a standard, established construct and must be developed and validated through empirical research.

Dimension Interface that foregrounds it The user’s implicit question

Object / Form Final preview, image, colour, form, WYSIWYG “What will it look like in the end?”

Spatial / Structure Map, hierarchy, flow, dependency, transformation “What is connected to what?”

Verbal / Explicit Description, rule, specification, label and explanation “Exactly what is defined and controlled?”

Outcome / Goal KPI, result, before/after, measurable output “What result will I get?”

The fourth row, Outcome, is not necessarily an independent, established cognitive style; this article proposes it as a practical dimension for product design. It is therefore important to distinguish “dimensions with theoretical support” from “exploratory dimensions extracted from data.”

Cognitive Style should not be equated with expertise either. Being a Designer does not mean being an Object Visualizer, and being a Developer does not mean being a Spatial Visualizer. A more realistic model is: Cognitive Style × Expertise × Goal × Context → Interface Response. Research must disentangle each contribution.

The central idea of this article—that one representation may not feel equally natural to all users, and that user behaviour can help select a more suitable representation—has a substantial research background. The article therefore does not claim to have “invented these concepts from scratch.” Its proposed value lies in bringing together several research strands developed at different times using different tools.

An important theoretical starting point is Vessey’s Cognitive Fit theory. Within this framework, graphs and tables can present the same information in cognitively different forms: graphs foreground spatial information, while tables foreground symbolic information. Performance improves when the representation, mental process and nature of the task align [19]. This theory shows that the “form of presentation” is not simply an aesthetic layer over information; it can directly affect problem-solving.

The next step in the literature was the automatic identification of cognitive differences from interaction itself. In 2007, Frias-Martinez, Chen and Liu showed that users’ interaction patterns with a digital library could be used to classify cognitive style automatically and personalise the interface accordingly. An important aim was to eliminate the need for lengthy psychometric tests on every occasion of use [20].

In 2009, Hauser, Urban, Liberali and Braun took the idea into dynamic web decision-making in “Website Morphing”: behaviour and clickstream data were used to rapidly update the probability that a user belonged to different cognitive styles, after which the site’s look & feel was adapted accordingly [21]. Conceptually, this is very close to the distinction this article draws between Functional Personalization and Cognitive Adaptation: functionality may remain constant while its presentation changes.

The closest precedent for Cognitive CAPTCHA is the 2012 paper by Papatheocharous and colleagues. In their study, 93 participants first chose between textual and graphical CAPTCHA and solved their preferred CAPTCHA; the choice was repeated ten times. Three categories of data—Preference, Processing Time and Ability/Attempts—were recorded and then fed into a neural network to predict the Verbal/Imager cognitive-style ratio. The authors explicitly stated that their ultimate aim was to use this prediction to create different interface experiences for different cognitive types [22]. Using CAPTCHA as a point for collecting cognitive signals therefore has a direct scientific precedent.

A subsequent study by Belk and colleagues in 2015 examined the relationship through two broader experiments. In the first study, involving 131 people, Verbals significantly preferred textual CAPTCHA and solved it faster, while Imagers showed a tendency to prefer and solve graphical CAPTCHA faster. In the second study, involving 125 people, processing speed, controlled attention and working memory capacity were also associated with performance on different CAPTCHA types. The authors concluded that cognitive differences should be considered when designing user-centred, personalised CAPTCHA [23].

Meanwhile, in 2012, Leung and colleagues directly examined the relationship between design preference and performance on cognitive tests by generating websites automatically. Participants experienced different information displays, including trees and tables, and selected their preferences; some relationships between cognitive abilities and representation preferences were observed [24]. This work is close to the third part of our proposed protocol: choosing between two interfaces with the same functionality but different representations.

The emergence of Generative UI now provides a dimension that was practically expensive or difficult in many of these earlier studies: rapidly producing multiple representations of the same functionality. In 2026, Peng and colleagues collected pairwise judgements from 20 designers on 600 generated interfaces, showing that design preferences vary substantially even among specialists and that an effective personalisation model can be built from lightweight, sparse feedback. Their proposed method outperformed baselines in generating personalised interfaces [25]. This study does not directly model Cognitive Style, but it shows that pairwise micro-feedback can become a practical mechanism for adapting interfaces in the Generative UI era.

Consequently, the innovation claimed here is not “the first adaptive interface,” “the first link between cognition and design,” or “the first use of CAPTCHA to infer cognition.” The distinctive proposal is to build a single cycle with very little friction: free Micro-Puzzle selection -> initial cognitive signal -> second adaptive selection/solution -> choice between two representations of the same functionality -> statistical learning of the relationship between the two -> and, ultimately, AI-assisted generation or selection of a suitable Cognitive Mode. Moreover, the proposed model need not classify people exclusively into predefined types from the outset. A theory-driven model such as Object/Spatial/Verbal can sit alongside a data-driven model of latent factors, allowing large-scale data to reveal a more realistic structure of the space.

The central challenge of this operational theory is that both sides of the equation are unknown. On one side, we do not know the user’s precise cognitive profile; on the other, we do not know which profile a particular design suits best. If we tell AI from the outset that “this interface is Object-oriented” and then use that same label to train the model, we have merely reproduced our own assumption.

The proposed solution is to shift from “top-down definition” to “behavioural measurement.” Instead of a lengthy questionnaire or a direct question such as “Are you a visual thinker?”, users encounter short, natural choices. The behaviour of large numbers of users can gradually clarify both users’ relative coordinates and interfaces’ relative coordinates.

In psychometrics, Forced Choice formats and multidimensional IRT models have been developed to infer latent traits from comparative choices [16]. More recent research shows that adaptive versions can select the next question using information from previous responses and achieve more precise estimates with fewer items [17]. The proposed Cognitive CAPTCHA draws inspiration from this literature, but it is not a valid psychometric test until its Item Bank has been calibrated.

Cognitive CAPTCHA is the provisional name used here for a three-stage interaction pattern. As the research background showed, using CAPTCHA and Preference, Processing Time and Ability data to predict cognitive style has already been tested [22,23]. The present design differs by treating “free choice of puzzle type” itself as the first signal, making the second stage an adaptive interaction between two Micro-Puzzles, and directly measuring preference between two representations of the same functionality in the third stage. The design principle is that the test should not become a personality questionnaire and that the successful path in the minimal version should require only three interactions; every click contributes both to interaction verification and to research data.

Click 1: freely choose 1 of 3 challenges Click 2: choose and solve 1 of 2 Micro-Puzzles Click 3: choose Interface A or B

Figure 3 - The minimal three-click version of Cognitive CAPTCHA; there is no additional confirmation button on the successful path.

Stage one: free choice

Three challenges are presented with difficulty, expected time, visual appeal and reward made as equal as possible, but with different representations: for example, form/image recognition, rotation and spatial relationships, and applying a short verbal rule. Users are asked to “solve whichever you prefer.” The choice itself is the first observation, not whether the answer is correct or incorrect.

Stage two: two adaptive puzzles

Based on the initial choice, two further Micro-Puzzles are shown. The user does not have to choose one and then press a start button; their first response simultaneously selects the puzzle and attempts to solve it. On the successful path, this stage requires one click. The types of puzzle in the second stage can be selected adaptively to maximise discriminatory power.

Stage three: the actual research question

Two interfaces or two representations of the same functionality are shown. Their functions should be as identical as possible, with only the presentation changing: for example, A could foreground the final result and preview, and B the structure and dependencies. The question is simple: “If you had to work with one every day, which would you choose?” This choice is the main target for learning the relationship between the cognitive signal and interface preference.

Recorded data Research use

Type of challenge selected Representation preference signal

Time to selection Relative strength or confidence of the choice

Correct response/error Distinguishing preference from performance

Switching puzzle or abandoning it Friction and dropout

Interface A/B choice Target variable for Cognitive Compatibility

Entry source and context Controlling channel, expertise and contextual effects

The hardest part of Cognitive CAPTCHA is not AI, but sound item design. If the Spatial puzzle is more attractive, easier or more familiar than the Verbal puzzle, the user’s choice may measure aesthetic preference or familiarity rather than cognitive processing. If one requires dragging and the other clicking, differences in motor interaction enter the data. If one option’s text is harder, literacy or language is measured instead of the intended construct.

• Options should be matched as closely as possible in difficulty, time, number of movements and visual quality.

• Option positions and left/right order should be randomised to reduce side bias.

• Anchor versions should be repeated to make response stability and random error measurable.

• Some users should also complete more established tests such as OSIVQ to calibrate the Item Bank.

• Preference and Cognitive Efficiency should be recorded separately; someone may choose the more attractive puzzle but be faster and more accurate on another.

• Items that cause high dropout, cultural bias or accessibility risk should be removed or revised.

In the initial phase, it is preferable to create hundreds of raw items, then retain only those with adequate discrimination, acceptable stability and minimal bias in the main bank. At this stage, AI can help generate variants, analyse patterns and select items adaptively, but Ground Truth comes from human behaviour and calibration tools.

To turn this idea into a scientific finding, we must move from an appealing claim to a reproducible protocol. The initial proposal can be implemented in several phases.

Phase Approximate sample Objective

Calibration 300 to 1,000 people Compare Micro-Puzzles with OSIVQ and remove weak items

Interface validation 1,000 to 5,000 people Test the relationship between the estimated profile and A/B choice, controlling expertise and context

Large scale 10,000 people and more Build a latent model, test cross-validation and examine subgroups

External validity Several independent recruitment sources Check that the model is not merely an artifact of one platform or reward incentive

Longitudinal testing Repeat after a week/month Measure profile stability and change

Testable hypotheses

H1. Initial selection among matched Micro-Puzzles will have a statistical relationship with scores on Object, Spatial and Verbal dimensions in more established instruments.

H2. Micro-Puzzle selection patterns can predict the probability of choosing certain interface representations better than a model without cognitive information.

H3. Response time and solution performance provide additional information beyond “choice of puzzle type.”

H4. Occupational expertise explains part of the variance in interface preference, but its effect will not fully replace cognitive dimensions.

H5. A model trained on one recruitment source will retain some of its predictive power in an independent source.

H6. If the same functionality is presented in a more compatible Cognitive Mode, continuation rate, task comprehension or satisfaction will improve relative to the fixed version; this hypothesis must be tested in a separate A/B test.

High Accuracy is not the only success criterion. The model must have appropriate calibration, confidence intervals, fairness across groups, robustness to changes in device and language, and reproducibility in an independent sample. Otherwise, percentages such as “90% suitable for Object” will merely create an appearance of scientific rigour.

If the relationship between cognitive patterns and interface representations is reproducible, the next step is not simply to score UI; the same engine can transform the interface. A single functionality can have multiple representations: Preview-first for someone who connects more quickly with the final result; Structure-first for someone who wants relationships and flow; and Specification-first for someone more comfortable with explicit rules and structured text.

Metaphorically, this resembles Dark Mode and Light Mode, but instead of changing brightness and colour, it changes how information is organised and presented. However, the system should not create a permanent label at the first interaction. A Cognitive Mode should be a tentative, revisable hypothesis, and the user should always be able to change it or return to the general display.

Conventional Personalization Proposed Cognitive Adaptation

What to show the user How to show the same thing

Based on role, need, history, permissions and behaviour Based on signals about processing and mental models

Example: a finance manager sees a financial Dashboard Example: the same Dashboard can centre on Preview, Flow or Specification

This approach therefore does not replace personalisation; it is an additional layer alongside it. Functional Personalization determines the appropriate content, while Cognitive Adaptation seeks to provide a more suitable representation of that same content.

Any method that uses small user behaviours to infer hidden characteristics can easily shift from a design tool to a profiling tool. Scientific and ethical validity must therefore be developed together.

• The results of a few clicks must not be presented as “personality diagnosis” or a definitive truth about an individual.

• In public-facing design, using data to improve or adapt the experience must be transparent and comply with consent and privacy requirements.

• Sensitive characteristics or high-risk decisions must not be inferred from such a profile.

• Accessibility must be guaranteed independently of the cognitive model; a user with visual, motor or linguistic limitations must not be misclassified because their performance differs.

• The model must allow change over time and across contexts; human cognition is not a fixed label.

• This approach must not be equated with simplified “Learning Styles” theories claiming that education should be adapted solely to self-reported preference. This article’s question concerns measurable prediction and interaction, not a presumed educational prescription.

It must also be acknowledged that preference does not always equal performance. An interface the user likes at first glance may produce lower productivity over prolonged use. First Impression, comprehension, task completion, cognitive load, trust and retention should therefore be measured separately.

The rapid growth of AI has created an exceptional opportunity to bring many longstanding theories and aspirations about cognitive differences out of research archives and turn them into practical, testable tools. The literature on Cognitive Fit, Website Morphing, automatic cognitive-style identification and CAPTCHA based on cognitive differences shows that many pieces of this puzzle have existed for years [19-24]. What has changed today is computational power, multimodal models, large-scale learning from behaviour and the possibility of generating different interfaces in real time with AI [25]. We may still be unable—and perhaps should not even attempt—to build a complete, definitive model of every person. But we can now design and produce different presentations of the same functionality for statistical groups with different cognitive patterns and measure their effects.

For a designer of management tools, this is an entirely practical question. If a few brief, unobtrusive steps—such as freely selecting a Micro-Puzzle, solving an adaptive puzzle and then choosing between two representations—can yield even a probabilistic signal of the user’s processing, AI can subsequently choose a presentation of the tool that increases willingness to continue, understanding or the sense that interaction feels natural. This method must not confine users to a definitive type; the model should be updated with every interaction.

This approach differs from conventional Personalization. Today’s tools can readily change according to the user’s needs, role and history. This article concerns “cognition,” rather than “need.” Two people may need exactly the same functionality, yet one understands it better on seeing the result, and the other on seeing the structure. In the proposed model, the system does not only ask “What do you need?” It asks “How do you understand that same thing better?”

When someone works in an environment whose representation is more compatible with their processing, they may feel more comfortable without knowing exactly why. If this hypothesis is confirmed in controlled experiments, digital experience quality will no longer depend only on aesthetics, features or speed; the degree of compatibility between the interface’s language and the user’s cognitive language will also become part of design.

The intelligent product of the future is not merely one that knows its user; it is one that can align its own way of understanding with the user’s way of understanding.

For me, this discussion does not end at the theoretical level. As a designer of management tools, I have spent years studying and practically exploring differences in how people understand tools, accept change and approach interfaces. Many ideas that previously remained unimplemented because of data collection costs, the difficulty of building multiple versions of an interface and modelling limitations have regained practical value with the emergence of the new generation of AI. I am now building and developing a working prototype of this tool with AI assistance and pursuing its delivery and testing on Trust-Login. There, a security decision can determine whether Cognitive CAPTCHA is displayed independently of the challenge’s content. If displayed, the challenge itself can both verify human interaction and collect research signals through a few short interactions. The first version’s goal is not to prove “personality diagnosis,” but to build infrastructure for gathering real data, calibrating puzzles and empirically testing the relationship between cognitive patterns and interface preference.

From this perspective, “taste” is not the end of an explanation, but the beginning of a question. Part of what we call taste may reflect the paths that perception, language, culture, experience and biology have taken in constructing each person’s mental world. Earlier research has shown that representation, cognitive style, digital behaviour and even CAPTCHA type can be related. Today’s AI allows these scattered pieces to be brought together within a single practical cycle. The proposed Cognitive CAPTCHA’s real value does not lie in claiming to understand a person in three clicks, but in building a tool with little friction to gather millions of observations, test hypotheses and turn the relationship between “how we understand” and “how we design” from intuition into a scientific, reproducible question that can ultimately be used in design.

The following references were used to distinguish established findings from this article’s hypotheses and proposals.

[1] Kozhevnikov, M., Evans, C., & Kosslyn, S. M. (2014). Cognitive Style as Environmentally Sensitive Individual Differences in Cognition: A Modern Synthesis and Applications in Education, Business, and Management. Psychological Science in the Public Interest, 15(1), 3-33. https://doi.org/10.1177/1529100614525555

[2] Blazhenkova, O., & Kozhevnikov, M. (2009). The new object-spatial-verbal cognitive style model: Theory and measurement. Applied Cognitive Psychology, 23(5), 638-663. https://doi.org/10.1002/acp.1473

[3] Kozhevnikov, M., Hegarty, M., & Mayer, R. E. (2002). Revising the Visualizer-Verbalizer Dimension: Evidence for Two Types of Visualizers. Cognition and Instruction, 20(1), 47-77. https://doi.org/10.1207/S1532690XCI2001_3

[4] Finn, E. S., Shen, X., Scheinost, D., et al. (2015). Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nature Neuroscience, 18, 1664-1671. https://doi.org/10.1038/nn.4135

[5] Schulz, M.-A., Baier, S., Timmermann, B., Bzdok, D., & Witt, K. (2021). A cognitive fingerprint in human random number generation. Scientific Reports, 11. https://doi.org/10.1038/s41598-021-98315-y

[6] Breit, M., Scherrer, V., Tucker-Drob, E. M., & Preckel, F. (2024). The Stability of Cognitive Abilities: A Meta-Analytic Review of Longitudinal Studies. Psychological Bulletin, 150(4), 399-439. https://doi.org/10.1037/bul0000425

[7] Winawer, J., Witthoft, N., Frank, M. C., Wu, L., Wade, A. R., & Boroditsky, L. (2007). Russian blues reveal effects of language on color discrimination. Proceedings of the National Academy of Sciences, 104(19), 7780-7785. https://doi.org/10.1073/pnas.0701644104

[8] Martinovic, J., Paramei, G. V., & MacInnes, W. J. (2020). Russian blues reveal the limits of language influencing colour discrimination. Cognition, 201, 104281. https://doi.org/10.1016/j.cognition.2020.104281

[9] Kearins, J. M. (1981). Visual spatial memory in Australian Aboriginal children of desert regions. Cognitive Psychology, 13(3), 434-460. https://doi.org/10.1016/0010-0285(81)90017-7

[10] Kearins, J. (1986). Visual spatial memory in Aboriginal and White Australian children. Australian Journal of Psychology, 38(3), 203-214. https://doi.org/10.1080/00049538608259009

[11] King, M. J., Katz, D. P., Thompson, L. A., & Macnamara, B. N. (2019). Genetic and environmental influences on spatial reasoning: A meta-analysis of twin studies. Intelligence, 73, 65-77. https://doi.org/10.1016/j.intell.2019.01.001

[12] Gero, J. S., & Kannengiesser, U. (2004). The situated function-behaviour-structure framework. Design Studies, 25(4), 373-391. https://doi.org/10.1016/j.destud.2003.10.010

[13] Dorst, K., & Cross, N. (2001). Creativity in the design process: co-evolution of problem-solution. Design Studies, 22(5), 425-437. https://doi.org/10.1016/S0142-694X(01)00009-6

[14] Jansson, D. G., & Smith, S. M. (1991). Design fixation. Design Studies, 12(1), 3-11. https://doi.org/10.1016/0142-694X(91)90003-F

[15] Lindgaard, G., Fernandes, G., Dudek, C., & Brown, J. (2006). Attention web designers: You have 50 milliseconds to make a good first impression! Behaviour & Information Technology, 25(2), 115-126. https://doi.org/10.1080/01449290500330448

[16] Brown, A., & Maydeu-Olivares, A. (2011). Item Response Modeling of Forced-Choice Questionnaires. Educational and Psychological Measurement, 71(3), 460-502. https://doi.org/10.1177/0013164410375112

[17] Qin, W., Zheng, Y., Kai, L., Yan, C., Siwei, P., & Dongbo, T. (2024). Item selection methods in multidimensional computerized adaptive testing for forced-choice items using Thurstonian IRT model. Behavior Research Methods, 56, 600-614. https://doi.org/10.3758/s13428-022-02037-6

[18] National Eye Institute. Types of Color Vision Deficiency. U.S. National Institutes of Health. https://www.nei.nih.gov/eye-health-information/eye-conditions-and-diseases/color-blindness/types-color-vision-deficiency

[19] Vessey, I. (1991). Cognitive Fit: A Theory-Based Analysis of the Graphs Versus Tables Literature. Decision Sciences, 22(2), 219-240. https://doi.org/10.1111/j.1540-5915.1991.tb00344.x

[20] Frias-Martinez, E., Chen, S. Y., & Liu, X. (2007). Automatic cognitive style identification of digital library users for personalization. Journal of the American Society for Information Science and Technology, 58(2), 237-251. https://doi.org/10.1002/asi.20477

[21] Hauser, J. R., Urban, G. L., Liberali, G., & Braun, M. (2009). Website Morphing. Marketing Science, 28(2), 202-223. https://doi.org/10.1287/mksc.1080.0459

[22] Papatheocharous, E., Belk, M., Germanakos, P., & Samaras, G. (2012). On Modelling Cognitive Styles of Users in Adaptive Interactive Systems using Artificial Neural Networks. In Proceedings of the 4th International Joint Conference on Computational Intelligence (NCTA-2012), 563-569. https://doi.org/10.5220/0004158905630569

[23] Belk, M., Fidas, C., Germanakos, P., & Samaras, G. (2015). Do human cognitive differences in information processing affect preference and performance of CAPTCHA? International Journal of Human-Computer Studies, 84, 1-18. https://doi.org/10.1016/j.ijhcs.2015.07.002

[24] Leung, S.-W., Lee, J., Johnson, C., & Robertson, D. (2012). Design preferences and cognitive styles: experimentation by automated website synthesis. Automated Experimentation, 4, 2. https://doi.org/10.1186/1759-4499-4-2

[25] Peng, Y.-H., Das, S., Bigham, J. P., & Wu, J. (2026). Efficient Personalization of Generative User Interfaces. arXiv:2604.09876. https://arxiv.org/abs/2604.09876