Skip to yearly menu bar Skip to main content


Poster
in
Workshop: Second Workshop on Representational Alignment (Re$^2$-Align)

Investigating the Role of Representation Switching Costs in Goal Persistence Bias

Gaia Molinaro · Aly Lidayan · Anne Collins


Abstract:

Goal pursuit profoundly shapes human cognition, typically benefiting learning and decision-making by focusing information processing. However, goal-dependent processing occasionally leads to seemingly maladaptive behavior, such as a bias toward goal persistence – the tendency to continue pursuing the current goal even when suboptimal. While various explanations have been proposed for such goal persistence, the underlying cognitive mechanisms remain unclear. We hypothesize that one key factor is the computational cost of switching between different internal representations of external stimuli that is induced by changes in the active goal. Humans' bias towards persisting with suboptimal goals could thus be resource-rational concerning their cognitive capacities. To test this, we developed a task where participants chose between competing goals, with rewards structured to encourage goal switching. Goals were organized in pairs requiring the same rules for action selection. In preliminary data (N = 29), participants showed both a bias towards goal persistence and a preference for switching between goals sharing the same rule. This preference correlated with individual differences in cognitive flexibility, i.e., a stronger preference was positively correlated with greater switch costs (i.e., increased reaction times) following different-rule switches compared to same-rule switches. Our preliminary results are consistent with our hypothesis that representation switch costs may play a role in goal persistence biases. In continuing the project, we will gather a larger sample size and recruit additional participants on modified versions of the task (N = 100 each). We predict that 1) we will replicate previous findings showing goal persistence; 2) goal switching will be biased toward same-rule goals; 3) this bias will correlate negatively with participants' cognitive flexibility; and 4) goal persistence biases will increase when all goals have different rules.

Chat is not available.