Association found between autistic traits and success at an exploration game

By Priyanjana Pramanik, MSc.Reviewed by Danielle Ellis, B.Sc.Oct 31 2024

Researchers tested 77 university students in a curiosity-driven exploration task. 

In a recent study published in PLoS Computational Biology, researchers explored how curiosity-driven behavior varies based on individual traits, particularly autistic traits, and its impact on exploration success.

Their findings highlight how individual differences in autistic traits shape exploration styles, with implications for the potential for personalized approaches to enhance learning processes.

Background

Curiosity-driven learning focuses on self-directed exploration, motivated by an intrinsic desire to learn rather than external rewards. People tend to explore environments where they expect to make more learning progress, disengaging when progress is minimal.

However, exploration behaviors vary significantly across individuals and may relate to personality traits like autistic traits, risk-taking, and impulsivity.

Autistic traits, including insistence on sameness, are associated with unique learning patterns, such as lower adaptability to uncertain or noisy situations. Past research shows those with higher autistic traits may exhibit less tolerance for prediction errors, affecting their exploration behaviors.

About the study

In this study, researchers explored how autistic traits affect curiosity-driven exploration. Their first hypothesis was that individuals displaying higher autistic traits may emphasize reducing uncertainty and value small, consistent learning progress. Alternatively, intolerance to uncertainty might lead individuals with high autistic traits to avoid situations with unpredictable outcomes.

Researchers recruited 77 participants who were either recent or current university students, of whom 70 continued into the study. The final participants were between 17 and 35, with an average age of 22.2; 14 identified as men, 51 as women, and 5 as non-binary.

Participants interacted with animal characters in a screen-based task, predicting each character's next location based on probabilistic hiding patterns. The task included three settings (grassland, sea, and beach), each with four animals.

The task allowed participants to explore freely, with choices tracked in relation to their prediction errors, learning progress, and novelty preferences. A hierarchical model assessed their trial-by-trial learning progress, prediction errors, and exploration choices. No instructions were provided, nor were rewards given if participants guessed correctly.

Additionally, researchers collected information on autistic traits through social behavior questionnaires designed for adults and, optionally, reports from participants' parents. The study focused on the "insistence on sameness" subscale, which evaluates the need for predictability and avoidance of change. Researchers also examined the broader impact that autistic traits may have on exploration behaviors.

By analyzing how autistic traits influence learning choices, the study aims to improve understanding of how these traits impact curiosity-driven exploration, differing between individuals.

Findings

Four logistic models tested the influence of factors (prediction error, learning progress, novelty) on participants' decisions to stay or leave. Autistic traits (especially "insistence on sameness") and time in trials were analyzed for their effects.

Participants with lower insistence on sameness used learning progress early on but switched to prediction error later. However, participants with higher insistence on sameness relied on learning progress later but did not use either factor initially. Novelty did not significantly impact these decisions.

Similar trends were observed when considering data from self-reports as explanatory variables, but not all interactions (particularly time) reached statistical significance.

On exploring the links between exploratory decisions and autistic traits, researchers found that participants with both high and low insistence on sameness preferred novel options.

Based on reports from others, novelty influenced both low and high insistence on sameness groups, while prediction error and learning progress effects were not significant. Based on self-reports, the low insistence group preferred options with lower prediction errors, while the high insistence group preferred options with higher learning progress.

In terms of associations with learning performance, higher insistence on sameness correlated with improved performance across most hiding patterns, except for a high-noise, unlearnable pattern. This interaction was significant with reports from others but not for self-reports.

Conclusions

Researchers examined how autistic traits affect curiosity-driven learning behaviors by using a task where participants chose when to stop sampling from an environment and what to explore next. They applied computational modeling to analyze participants' learning progress and prediction errors.

While participants with lower insistence on sameness relied more on learning progress to leave an environment early on, they switched to using expected prediction error to leave activities if they anticipated poor performance.

Participants with higher insistence on sameness showed greater persistence, relying less on learning progress initially but gradually started leaving activities only if learning progress decreased. All participants preferred novel options.

However, other autistic traits, such as reduced social interaction and empathy, may also influence exploration beyond insistence on sameness. Researchers highlighted the need for future research to explore brain mechanisms and causal links between autistic traits and learning behaviors.