As early as 1949, Weaver defined communication in a very broad sense to include all procedures by which one mind or technical system can influence another, thus establishing the idea of semantic communication. With the recent success of machine learning in expert assistance systems where sensed information is wirelessly provided to a human to assist task execution, the need to design effective and efficient communications has become increasingly apparent. In particular, semantic communication aims to convey the meaning behind the sensed information relevant for Human Decision-Making (HDM). Regarding the interplay between semantic communication and HDM, many questions remain, such as how to model the entire end-to-end sensing-decision-making process, how to design semantic communication for the HDM and which information should be provided for HDM. To address these questions, we propose to integrate semantic communication and HDM into one probabilistic end-to-end sensing-decision framework that bridges communications and psychology. In our interdisciplinary framework, we model the human through a HDM process, allowing us to explore how feature extraction from semantic communication can best support HDM both in theory and in simulations. In this sense, our study reveals the fundamental design trade-off between maximizing the relevant semantic information and matching the cognitive capabilities of the HDM model. Our initial analysis shows how semantic communication can balance the level of detail with human cognitive capabilities while demanding less bandwidth, power, and latency.

In decision-making scenarios, individuals often face the challenge of balancing between exploring new options and exploiting known ones-a dynamic known as the exploration-exploitation trade-off. In such situations, people frequently have the opportunity to observe others' actions. Yet little is known about when, how, and from whom individuals use observational learning in the exploration-exploitation dilemma. In two experiments, participants completed multiple nine-armed bandit tasks, either independently or while observing a fictitious agent using either an explorative or equally successful exploitative strategy. To analyze participants' behaviors, we used a reinforcement learning model (simplified Kalman Filter) to extract parameters for both copying and exploration at the individual level. Results showed that participants copied the observed agents' choices by adding a bonus to the individually estimated value of the observed action. While most participants appear to use an unconditional copying approach, a subset of participants adopted a copy-when-uncertain approach, that is copying more when uncertain about the optimal action based on their individually acquired knowledge. Further, participants adjusted their exploration strategies in alignment with those observed. We discuss, in how far this can be understood as a form of emulation. Results on participants' preferences to copy from explorative versus exploitative agents are ambiguous. Contrary to expectations, similarity or dissimilarity between participants' and agents' exploration tendencies had no impact on observational learning. These results shed light on humans' processing of social and non-social information in exploration scenarios and conditions of observational learning.

Sequential decision-making is a common cognitive task where subsequent decisions often depend on the outcomes of earlier ones. While sequence learning research demonstrates humans' ability to learn regularities in sequentially presented information, investigations are sparse regarding complex decision-making tasks, such as category learning. This study connects both domains and explores whether individuals can detect and utilize regularities between sequential categorization task outcomes to enhance learning and categorize novel targets. For this, we extended a classical category learning paradigm (Study 1: Type I, Study 2: Type II category structures), where the outcome of one categorization task depends on the outcome(s) of previous tasks in a sequence. We compared performance in each study to a control condition without dependencies (Type VI). Connecting the design to sequential grammar learning, in Study 1, we further manipulated the adjacency of the relevant outcomes (consecutive or separated by an irrelevant task).The results of Study 1 showed that with a Type I dependency, participants learned the second task's outcome more rapidly than in the control condition. During the transfer phase, participants successfully applied the dependency to categorize novel targets in both adjacent and non-adjacent conditions. In contrast, in Study 2, we found no evidence of effects on learning or generalization of a Type II dependency, as performance was equal to the control condition. We discuss these findings from category learning and statistical learning perspectives and how investigations intersecting both domains can contribute to the broader understanding of complex sequential decision-making processes. We also highlight open questions for future research.

Previous research has shown that human participants performed suboptimally in patch-leaving behavior during foraging tasks. This suboptimal performance stemmed from two primary sources: Participants often adopted a strategy unsuited to the environment and failed to apply it optimally. The current study investigates whether providing feedback on participants' patch-leaving behavior can improve their performance by facilitating either a switch to a more effective strategy or an enhanced application of their existing strategy. All participants completed a patch-leaving task across three sessions: pre-feedback, feedback, and post-feedback. Their patch-leaving strategies in each session were identified through computational modeling. During the feedback session, participants received feedback based on either the fixed-time (FT) or giving-up-time (GUT) strategy. Most participants employed the GUT strategy in the pre-feedback session and showed improved performance in the post-feedback session. In the FT feedback condition, many participants switched to using the FT strategy in the post-feedback session. Participants who switched improved in performance, whereas those who continued using the GUT strategy did not. In contrast, in the GUT feedback condition, most participants continued using the GUT strategy but benefited from feedback due to a more precise execution of the GUT strategy in the post-feedback session. These results suggest that participants can adapt to a better-suited strategy or improve their application of a suboptimal strategy with appropriate feedback.

Classification is a common cognitive task, which requires assigning objects or events to categories based on shared features or rules (e.g., red objects are fruit, brown objects are mushrooms). In everyday scenarios, however, objects usually belong to more than one category (e.g., red objects can also be classified as edible, and brown objects could be poisonous). This study investigates whether humans can learn corresponding regularities between outcomes of such multiple categorizations when performed in a series of decisions for each stimulus. We therefore translated classical category learning designs, known as Type I (one-dimensional rule) and Type II (disjunctive rule), into a temporal context. We compared these cases to conditions in which no correlations existed between the series of categorization outcomes, and only the visual stimulus predicted each category outcome. Besides the structural complexity, we also tested in Type I scenarios whether learning and generalization were moderated by the temporal proximity of the successive decisions (adjacent vs. non-adjacent categorizations). The results show that participants can abstract away from the visual stimulus with a temporal Type I regularity, but there was no evidence for a corresponding effect with a temporal Type II regularity. The role of adjacency was not clear-cut, but there was no strong evidence favoring stronger performance with adjacent relative to non-adjacent categorizations. We discuss these findings before the background of category- and artificial grammar-learning research, and expand on potential moderating factors such as the cognitive effort of keeping the necessary amount of information in working memory and the modality of category predictors when determining whether people will extract rules or rely on memory-based learning.

Quantitative judgments have been suggested to result from a mixture of similarity- and rule-based processing. People can judge an object’s criterion value based on the object’s similarity to previously experienced exemplars and based on a rule that integrates the object’s cues like a linear regression. In order to better understand these processes, the present work combines cognitive modeling and eye tracking and tests whether people who rely more on the similarity to exemplars also look more at the exemplar locations on the screen. In two eye-tracking studies, participants learned to assign each of four exemplars to a different screen corner and criterion value and then judged the criterion value of briefly presented test stimuli. Eye tracking measured participants’ gazes to the now empty exemplar locations (a phenomenon called looking-at-nothing); cognitive modeling of the test phase judgments quantified participants’ reliance on a similarity- over a rule-based process. Participants showed more similarity use and more looking-at-nothing in the study in which the cues were linked to the criterion by a multiplicative function than in the study with an additive cue-criterion link. Focusing on the study with a multiplicative environment, participants relying more on the similarity to exemplars also showed more looking-at-nothing ( τ = 0.25, p = .01). Within trials, looking-at-nothing was usually directed at the one exemplar that was most similar to the test stimulus. These results show that a multi-method approach combining process tracing and cognitive modeling can provide mutually supportive insights into the processes underlying higher-order cognition.

Previous research has shown that transitions in reward prospect influence (voluntary) task switching behavior. Specifically, an increase in reward prospect appears to enhance flexibility, as indicated by a higher voluntary switch rate (VSR), compared to situations where the reward prospect remains high. In contrast, when participants are randomly rewarded in the previous task, they tend to stick with this task, resulting in a lower VSR. The present study further explores the impact of probabilistic reward schemes on task switching. Two tasks were associated with distinct probabilities of receiving a reward for correct responses (high vs. low probability). This design allows for more refined predictions regarding VSR based on the results summarized above. In three experiments with voluntary and cued task switching, we observed that participants switched tasks less frequently when they were rewarded on the previous trial, regardless of whether the task had a high or low reward probability. This pattern suggests the use of a win-stay, lose-shift (WSLS) strategy, where participants are more likely to repeat their choice after receiving a reward. However, reward had no impact on switch costs. These results are discussed in the broader context of decision-making research, particularly in relation to strategies like WSLS and possibly different levels of cognitive processes affected by our manipulation and that of studies investigating transitions of reward prospect.

People often categorize the same object variably over time. Such intraindividual behavioral variability is difficult to identify because it can be confused with a bias and can originate in different categorization steps. The current work discusses possible sources of behavioral variability in categorization, focusing on perceptual and cognitive processes, and reports a simulation with a similarity-based categorization model to disentangle these sources. The simulation showed that noise during perceptual or cognitive processes led to considerable misestimations of a response determinism parameter. Category responses could not identify the source of the behavioral variability because different forms of noise led to similar response patterns. However, continuous model predictions could identify the noise: Noisy feature perception led to variable predictions for central stimuli on the category boundary, noisy feature attention increased the prediction variability for stimuli differing from each category on another feature, and noisy similarity computation increased the variability for stimuli with moderate predictions. Measuring category beliefs in a continuous way (e.g., through category probability judgments) may therefore help to disentangle perceptual and process-related sources of behavioral variability. Ultimately, this can inform interventions aimed at improving human categorizations (e.g., diagnosis training) by indicating which steps of the categorization mechanism to target.

Algorithmic advice has the potential to significantly improve human decision-making, especially in dynamic and complex tasks that require a balance between exploration and exploitation. This study examines conditions under which individuals are willing to accept advice from algorithms in such scenarios, focusing on the interaction between participants' exploration preferences and those of the advising algorithm. In an online experiment, we designed reinforcement learning algorithms to prioritize either exploration or exploitation and observed participants' decision-making behavior, modeled using a cognitive framework analogous to the algorithm. Contrary to expectations, participants did not show a preference for algorithms that matched their own exploration tendencies. In particular, participants were more likely to follow the advice of exploitative, consistent algorithms, possibly interpreting consistency as an indicator of competence. Although the participants benefited from the advice of the exploratory algorithm, their reluctance to follow it, regardless of whether the recommendation had been ignored previously or not, highlights a potential challenge in promoting effective collaboration between humans and algorithms. Explorative algorithms have the potential to promote behavioral diversification, but this effect is negated when humans disregard their advice. In such cases, algorithmic guidance can unintentionally decrease behavioral diversity by reinforcing established patterns.

Category learning is essential for making sense of the complex world around us. Unlike traditional laboratory settings, real-world learning often allows individuals to self-regulate their learning process, deciding when they have acquired sufficient knowledge to differentiate between categories. This study investigates how category variability-the extent to which exemplars within a category differ-shapes the duration of the learning process in a novel self-regulated task. Participants explored exemplars from two categories, determining for themselves when they had learned enough to categorize accurately. We found that increased variability within the focal category led participants to sample more extensively, suggesting that learners weigh the costs of continued exploration against the uncertainty introduced by environmental demands. Additionally, the variability of the counter-category emerged as a significant factor influencing the search and learning process, underscoring the relational nature of category acquisition. By examining the interplay between variability in both focal and counter-categories, this study provides novel insights into how learners acquire categories and effectively regulate their learning in response to variability.

Sequential decision-making, where choices are made one after the other, is an important aspect of our daily lives. For example, when searching for a job, an apartment, or deciding when to buy or sell a stock, people often have to make decisions without knowing what future opportunities might arise. These situations, which are known as optimal stopping problems, involve a risk associated with the decision to either stop or continue searching. However, previous research has not consistently found a clear connection between individuals' search behavior in these tasks and their risk preferences as measured in controlled experimental settings. In this paper, we explore how particular characteristics of optimal stopping tasks affect people's choices, extending beyond their stable risk preferences. We find that (1) the way the underlying sampling distribution is presented (whether it is based on experience or description), (2) the sequential presentation of options, and (3) the unequal frequencies of choices to reject versus to accept significantly bias people choices. These results shed light on the complex nature of decisions that unfold sequentially and emphasize the importance of incorporating context factors when studying human decision behavior.

The retrieval of past instances stored in memory can guide inferential choices and judgments. Yet, little process-level evidence exists that would allow a similar conclusion for preferential judgments. Recent research suggests that eye movements can trace information search in memory. During retrieval, people gaze at spatial locations associated with relevant information, even if the information is no longer present (the so-called 'looking-at-nothing' behavior). We examined eye movements based on the looking-at-nothing behavior to explore memory retrieval in inferential and preferential judgments. In Experiment 1, participants assessed their preference for smoothies with different ingredients, while the other half gauged another person's preference. In Experiment 2, all participants made preferential judgments with or without instructions to respond as consistently as possible. People looked at exemplar locations in both inferential and preferential judgments, and both with and without consistency instructions. Eye movements to similar training exemplars predicted test judgments but not eye movements to dissimilar exemplars. These results suggest that people retrieve exemplar information in preferential judgments but that retrieval processes are not the sole determinant of judgments.

Category variability or diversity is an important factor influencing generalisation. However, expectations of category variability may not only depend on the variability of encountered category members, but may also be shaped by prior experiences with similar categories. In this study, we investigated whether we could influence category generalisation by inducing different category representations in an A/Non-A categorisation task: Participants either learned about a homogeneous category Non-A or a diverse category Non-A during a priming phase. To better understand the transfer process, we varied the nature of the learning phase from implicit transfer to explicit instructions that actively requested participants to use their prior experiences. We found that while with a homogeneous Non-A representation, generalisation of the A and Non-A categories was equal, the generalisation of category Non-A widened after a priming phase with a diverse representation. In a second experiment, we found that the widening of generalisation of category Non-A occurred when the exemplars in this category were themselves diverse (feature-diverse condition) but not when the category contained distinct exemplars (exemplar-diverse condition). These results suggests that categorisation is influenced by previous categorisation experiences possibly altering the representation of a category. Furthermore, the study gives a hint what kind of heterogeneity is needed to observe the commonly reported broader generalisation of diverse categories. The finding has implications not only to understand the influence of prior experiences on category learning, but any cognitive process that hinges on generalisation.

Stimulus classification is an everyday feat (e.g., in medical diagnoses by differentiating ultrasound images). Category feedback, however, is often non-deterministic (e.g., by 25% chance untrue a.k.a. probabilistic feedback) rendering experiences as somewhat unreliable. In probability learning and economic decisions (when humans try to predict which of two outcomes is more rewarding), it is often observed that humans decide non-rational (probability matching, Gamblers Fallacy). Despite shared origins, the research areas of category learning, probability learning, conditioning and economic decisions (experience based) did not arrive at a consensus of what drives such probability matching strategies. Here, we offer a domain-general integrative model that predicts the mixture of behavioral trends when participants learn probabilistic stimulus-outcome regularities. We use the Category Abstraction Learning framework (CAL), implementing the hypothesis, that humans count the streak of events, and generate simple rules and conditional hypotheses with them. We present simulations of four studies, one from each domain, showing that CAL’s learning mechanisms accurately predict systematic and individual differences in category (correlation) learning, reward learning, risky gambles, and fear conditioning, concerning the phenomena of Gamblers Fallacy, positive and negative recency, or win-stay-lose-shift strategies, in a single model. One central novel CAL hypothesis to link learning phenomena under gains, losses and neutral feedback, is that gains and losses differently affect cognitive control during learning (corrective feedback processing), thereby also providing a novel perspective on risk-preferences in experience-based risky gambles. We discuss CAL’s potential as a domain-general account of human learning in experience-based decisions in light of a broader range of theories from multiple domains.

Multi-criteria decision analysis (MCDA) is well suited to address complex public policy problems but could benefit from new tools to involve many laypeople. Online information on specialized topics could be more engaging by including game elements. This paper reports an experiment that assessed a gamified interface to (1) inform laypeople about the objectives to consider in wastewater management decisions, (2) assist them in constructing range-based preferences, and (3) provide a positive experience. We measured the effects with (1) a knowledge pre- and posttest, (2) the elicited weights and a range sensitivity index, and (3) an experience questionnaire based on self-determination theory. Answers from 174 participants indicated that participants learnt about the objectives and constructed preferences in both the gamified and control treatments. However, in neither were weights sufficiently adjusted. Our gamification making the ranges salient did not help overcome this bias. Both treatments were experienced as neutral to positive, the gamified being more entertaining. We discuss implications: if gamification of tools for participatory decision-making is to be promoted, it requires further research. Range insensitivity remains an unresolved bias in MCDA.