# Asymmetric learning and adaptability to changes in relational structure during transitive inference

## TL;DR

Humans update beliefs more about winners than losers, which helps infer relationships but can hinder adapting when high-ranking items drop in rank.

## Contribution

The study shows how winner-biased learning affects adaptability to changes in relational structures during transitive inference.

## Key findings

- Participants learned better when a low-ranking item rose in rank than when a high-ranking item dropped.
- A reinforcement learning model captured participants' winner-biased learning and adaptability.
- Well-performing participants could reduce or reverse their winner bias to adapt to changes.

## Abstract

Humans and other animals can generalise from local to global relationships in a transitive manner. Recent research has shown that asymmetrically biased learning, where beliefs about only the winners (or losers) of local comparisons are updated, is well-suited for inferring relational structures from sparse feedback. However, less is known about how belief-updating biases intersect with humans’ capacity to adapt to changes in relational structure, where re-valuing an item may have downstream implications for inferential knowledge pertaining to unchanged items. We designed a transitive inference paradigm involving one of two possible changepoints for which an asymmetric (winner- or loser-biased) learning rule was more or less optimal. Participants (N = 83) exhibited differential sensitivity to changes in relational structure: whereas participants readily learned that a hitherto low-ranking item increased its rank (‘up’ condition), moving a high-ranking item down the hierarchy impaired downstream inferential knowledge (‘down’ condition). Behaviour was best captured by a reinforcement learning model which exhibited an initially winner-biased learning strategy that was nonetheless adaptable – that is, while this winner bias predominantly limited participants’ flexibility in the ‘down’ condition, well-performing participants were able to reduce or even reverse their winner bias in order to appropriately accommodate the relational change. Our results indicate that asymmetric learning not only accounts for efficient inference of latent relational structures but also for differences in the ease with which learners accommodate structural changes.

Human relational learning is winner-biased: we update our beliefs about winners more than losers. A transitive inference study with computational modelling shows this bias hinders flexibility when adapting to certain relational changes over others.