(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Autistic traits foster effective curiosity-driven exploration [1]

['Francesco Poli', 'Donders Institute For Brain', 'Cognition', 'Behaviour', 'Radboud University', 'Nijmegen', 'Mrc Cognition', 'Brain Sciences Unit', 'University Of Cambridge', 'Cambridge']

Date: 2024-12

Abstract Curiosity-driven exploration involves actively engaging with the environment to learn from it. Here, we hypothesize that the cognitive mechanisms underlying exploratory behavior may differ across individuals depending on personal characteristics such as autistic traits. In turn, this variability might influence successful exploration. To investigate this, we collected self- and other-reports of autistic traits from university students, and tested them in an exploration task in which participants could learn the hiding patterns of multiple characters. Participants’ prediction errors and learning progress (i.e., the decrease in prediction error) on the task were tracked with a hierarchical delta-rule model. Crucially, participants could freely decide when to disengage from a character and what to explore next. We examined whether autistic traits modulated the relation of prediction errors and learning progress with exploration. We found that participants with lower scores on other-reports of insistence-on-sameness and general autistic traits were less persistent, primarily relying on learning progress during the initial stages of exploration. Conversely, participants with higher scores were more persistent and relied on learning progress in later phases of exploration, resulting in better performance in the task. This research advances our understanding of the interplay between autistic traits and exploration drives, emphasizing the importance of individual traits in learning processes and highlighting the need for personalized learning approaches.

Author summary Research has long recognized that individuals display curiosity and explore their environments in order to learn. It is suggested that personal characteristics, including autistic traits, might influence how one engages in such exploratory behaviors. In this study, participants with varying levels of autistic traits participated in a game of locating hidden characters. We aimed to understand their decision-making process: which character they decided to engage with and for how long. Remarkably, participants with stronger autistic traits exhibited distinct exploration patterns, and in scenarios requiring persistence, their approach was particularly effective. This research underscores the importance of recognizing that individuals, especially those with autistic traits, may possess unique strategies for exploration and learning. This realization can guide educators and policy-makers in crafting more tailored learning environments. Furthermore, it emphasizes that the presence of autistic traits can be associated with specific strengths, reshaping our understanding and appreciation of neurodiversity.

Citation: Poli F, Koolen M, Velázquez-Vargas CA, Ramos-Sanchez J, Meyer M, Mars RB, et al. (2024) Autistic traits foster effective curiosity-driven exploration. PLoS Comput Biol 20(10): e1012453. https://doi.org/10.1371/journal.pcbi.1012453 Editor: Alireza Soltani, Dartmouth College, UNITED STATES OF AMERICA Received: October 9, 2023; Accepted: September 3, 2024; Published: October 31, 2024 Copyright: © 2024 Poli et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: Data and code for computational models and statistical analyses are available on OSF: https://osf.io/h2prm/ (DOI 10.17605/OSF.IO/H2PRM). Funding: This study was supported by a Donders Centre for Cognition internal grant to S.H. and R.B.M. (“Here’s looking at you, kid.” A model-based approach to interindividual differences in infants’ looking behavior and their relationship with cognitive performance and IQ; award/start date: 15 March 2018), a VICI grant from the Netherland Organization for Scientific Research NWO to S.H. (“Loving to learn - How curiosity drives cognitive development in young children”; serial number: VI.C.191.022), a Wellcome Trust center grant to benefit of R.B.M. (“Wellcome Centre for Integrative Neuroimaging”; serial number: 203139/Z/16/Z), a EPA Cephalosporin Fund and Biotechnology and Biological Sciences Research Council to R.B.M. (BB/N019814/1). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

1. Introduction How agents define their own learning curriculum, actively selecting what they want to explore, plays a pivotal role in learning [1]. This aspect of exploration stems from an intrinsic drive to learn, rather than from maximizing any extrinsic rewards. For this reason, it is often called curiosity-driven [2–3]. Recent evidence shows that humans explore different environments and engage in different activities taking into account the learning opportunities they offer [4–6]. Specifically, it was shown that participants playing a learning game were more likely to stop exploring an environment when they were making only little learning progress. Moreover, they were more likely to select activities in which they expected to make more learning progress [4]. The relationship between individual differences in curiosity-driven exploration and aspects of personality and real-world behaviors remains largely unexplored. Recent research has shown that exploration levels are highly variable across individuals [7], and possibly related to personality traits such as impulsivity and risk-taking [7,8]. This provides initial evidence that the mechanisms underlying exploratory behavior may be influenced by individual differences in personality traits. In this perspective, the exact mechanisms that drive exploratory behavior may be confounded at the group level [9], and can only be properly identified and understood by accounting for individual differences in personality traits. To explore this, we tested how the cognitive mechanisms underlying curiosity-driven exploration related to autistic traits, which represent a significant source of variability in personality traits within the general population [10,11]. Although the sensory [12–14], cognitive [15–17], social [18,19] and communicative aspects [20–22] of autistic traits have been studied thoroughly, recent theoretical work and empirical findings have sparked interest in how autistic traits relate to individual differences in learning mechanisms [23,24]. It was found that stronger autistic traits were linked to less efficient learning about probabilistically aberrant events [25,26], and the learning of participants with high insistence on sameness was less robust to noise [27]. However, autistic and non-autistic participants performed in the same way in visual search and decision-making tasks [28,29], and when environments were volatile [30], and some studies also showed enhanced statistical learning abilities in autistic people [31]. None of these studies focuses on the active aspect of learning, which has received surprisingly little attention in relation to autistic traits. A recent study [32] investigated how, during exploration, autistic traits relate to differences in the tolerance to prediction errors (i.e., the mismatch between expected outcome and actual outcome). Participants could freely move between two environments, while they had to discover their latent structure. Participants with fewer autistic traits had a greater tolerance for prediction errors before opting to abandon the environment. These findings are consistent with the fact that autistic traits have been linked to insistence on sameness, intolerance to errors, and difficulty dealing with sudden changes [33–35]. Yet, it remains unknown how these traits might alter the balance between different curiosity drives, such as seeking out novelty [36], minimizing prediction errors [37], and maximizing learning progress [38]. The lack of tolerance for uncertainty may influence curiosity-driven exploration behavior in two different ways. The first possibility is that individuals with higher autistic traits scores may be more motivated to reduce uncertainty. Given that learning progress allows uncertainty to reduce, they might thus give more importance to learning progress, even when it comes in small amounts. This predicts a stronger relationship between learning progress and exploration for people with higher scores on autistic traits. A second possibility is that the intolerance to uncertainty may lead to avoiding uncertain situations as much as possible [35,39]. As a consequence, people who score higher on autistic traits might actively avoid prediction errors. From this follows that their exploratory decisions would be guided by the avoidance of prediction errors rather than an interest in learning progress. To assess these hypotheses, we tested participants on a learning task in which they could freely interact with different animal characters on a screen. The animals appeared following probabilistic patterns and participants were instructed to predict the next location of the animal. We generated these hiding patterns so that participants could learn about the likely next hiding location of the animal and improve their prediction performance (i.e., make learning progress). We manipulated various sources of uncertainty to generate learnable situations that allowed measuring participants’ predictions and learning. Crucially, we monitored when participants decided to stop exploring an animal’s hiding pattern, which animal they picked next, and whether this depended on their learning progress, their prediction errors, or the novelty of the animals. To better assess their performance and exploratory drives, we fit participants’ behavioral responses to a hierarchical delta-rule model [4]. This allowed us to obtain trial-by-trial estimates of each participant’s prediction errors and learning progress on the task, as well as their expectations of how much learning progress and prediction error they might experience in future trials. We examined how exploration drives are affected by autistic traits through relating trial-by-trial estimates of (expected) learning progress and prediction error to exploration choices and self- and other-reports of autistic traits. Given that the need for predictability might be the reason why some people entertain repetitive behaviors and have specialized interests, the primary focus of our analyses was on the subscale of insistence on sameness, which assesses these aspects [40,41]. However, the same predictions may hold for multiple subscales, or for the total scale for autistic traits, because multiple traits may influence exploration decisions in similar ways. Therefore, we also explored the possibility of general effects of autistic traits on exploration behavior, in addition to examining the subscale of insistence on sameness. By investigating these possibilities, we aim to gain a better understanding of how autistic traits affect curiosity-driven exploration and how individuals with higher autistic traits scores might differ in their exploration drives from individuals with lower scores.

4. Discussion To examine how curiosity-driven exploration relates to autistic traits, we tested participants in a learning task in which they could actively decide when to stop sampling from an environment and what to explore next. Through computational modelling, we obtained trial-by-trial estimates of participants’ prediction errors and learning progress, as well as of their expectations about future errors and progress. This allowed us to examine which of the factors that have previously been suggested to drive exploration behaviors [4,36,37] determine exploration and learning in people with different degrees of autistic traits. To relate these drives to autistic traits, we collected information about autistic traits in a sample of university students using the ASBQ questionnaire [43], both through other- and self-reports. We found that autistic traits affect curiosity-driven behavior, in terms of leave-stay and exploratory decisions, as well as learning outcomes. Concerning leave-stay decisions, participants who scored lower on insistence on sameness were more likely to leave an environment if their learning progress was lower during the initial phase of exploration. Hence, they initially relied more on learning progress than participants who scored higher on insistence on sameness. However, during later sampling they stopped relying on learning progress, and exploited expected prediction error instead: They were more likely to leave an activity if they expected to perform badly in the near future. In participants with more insistence on sameness, this pattern of results was different. Participants relied less on learning progress and expected prediction error at the start of the task, and were simply less likely to abandon an environment, showing greater persistence. Only later in time, they started to rely on learning progress, abandoning activities that offered only relatively little learning progress. When deciding what to explore next, participants preferred more novel options, irrespective of their autistic traits. Yet, participants with lower self-reports of insistence on sameness additionally relied on the expected prediction errors, that is, they preferred options on which they expected to perform better. Conversely, participants with higher self-reports of insistence on sameness relied on the expected learning progress of the available options: They picked the option that offered a greater opportunity to learn. This result can be interpreted in terms of differential utility functions for people with lower and higher insistence on sameness, where the former tried to avoid errors, while the latter tried to maximize learning. These different drives were related to learning performance. Higher insistence on sameness scores were related to better performance on more probabilistic patterns. This better performance was also observed for the more complex high-drift sequences. Hence, high insistence on sameness is advantageous when tackling complex tasks that require more persistence. Overall, this pattern of results supports our initial hypothesis that insistence on sameness relates to an increased sensitivity to learning progress, which shows in the current task as increased reliance on learning progress to make leave-stay decisions and increased reliance on expected learning progress when deciding what to explore next. The ability to flexibly adjust to changes in the environments is crucial for effective learning across life [51–53] and it has been found to be reduced in individuals with high autistic traits [26] and in individuals with high attention to detail [27]. This was seen as a form of learning impairment, namely the inability to integrate data to overcome environmental noise [27]. On the contrary, in our task, we observe improvements in performance due to increased insistence on sameness and overall autistic traits. This may be explained by the fact that, in contrast to previous research, we let participants free to choose the activities they want to learn from, and to learn at their own pace. This might have been especially beneficial for individuals who scored high on insistence on sameness, because their natural inclination to avoid change allowed them to persevere on the task. Conversely, low insistence on sameness might be detrimental in this context, as it might lead individuals to flit from one activity to another without delving deep into any of them. The pattern of exploratory behaviors was modulated by insistence on sameness in a similar way as by the full scale of autistic traits, as well as some of the subscales, such as reduced contact and reduced empathy for leave-stay decisions, and reduced contact and reduced social interactions for exploratory decisions. These additional effects suggest that the relation between autistic traits and exploration drives is not confined to insistence on sameness, as it extends to a broader range of autistic traits. This result might be due to a causal relation between the different autistic traits. For example, higher insistence on sameness might lead to reduced social interactions [54]. These relations between traits have been found to be stronger in females than males [55], which may explain why we also observe them in our sample, that was in majority composed of female participants. Alternatively, broader genetic [56], chemical [57,58], or brain-related [59,60] factors might impact multiple traits at the same time. Future research should examine the exact brain mechanisms that affect both autistic traits and exploratory behavior, and the causal links between them. When studying the active aspects of learning, researchers need to find ways of handing control over to the participants without losing experimental rigor. This makes this type of research different from other areas of cognitive science, where paradigms are mostly passive, and participants are required to follow specific instructions (e.g., [61]). Although the active approach comes with unique benefits (i.e., studying aspects of behavior that are driven by intrinsic motivation and engagement), it also requires more data curation. The lack of precise instructions led us to exclude several participants, and more trials were excluded depending on participants’ boredom. Future studies should include bigger samples to assess how intrinsically motivated behaviors change as a function of these preprocessing decisions. Finally, when analyzing exploratory decisions only, we were constrained by our analytic tools to split the sample in two groups, and look at each subgroup’s results independently. This resulted in a loss of granularity in the analysis of exploratory decisions compared to our analysis of leave-stay decisions. Future studies are needed to develop analytic methods to examine traits as continuous even in multinomial models.

5. Conclusions In this study, we provide insights into how autistic traits influence exploration behavior. By using a novel exploration task and a hierarchical delta-rule model, we show that individuals with lower insistence-on-sameness scores are more likely to rely on learning progress early on but switch to expected prediction error later. Instead, individuals with higher scores on the same subscale are less likely to abandon an activity, showing greater persistence, and eventually start to rely on learning progress. We also show that individuals with lower insistence on sameness are more likely to choose options with low expected prediction error, while those with higher insistence on sameness tend to select options with high expected learning progress. These findings provide a nuanced understanding of how autistic traits influence exploration behavior and learning outcomes. Shifting the focus from what individuals can do to what they are motivated to do, we show how learning abilities change across different degrees of autistic traits.

Supporting information S1 Text. Supplementary tables, figures, and analyses. Fig A. The preprocessing pipeline. Fig B. Participants’ predictions. Fig C. Model simulations. Fig D. Parameters recovery. Fig E. Model comparison. Fig F. Comparing behavioral data to the predictions of the logistic model. Table A. Relation between leave-stay decisions and other-reports. Table B. Relation between leave-stay decisions and self-reports. Appendix A. Task-related questionnaire. Appendix B. Supplementary analyses. https://doi.org/10.1371/journal.pcbi.1012453.s001 (DOCX)

[END]
---
[1] Url: https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1012453

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.