# Information uncertainty influences learning strategy from sequentially delayed rewards

## TL;DR

The paper explores how people assign credit to past actions when rewards are delayed and information is uncertain.

## Contribution

It introduces two learning strategies and shows how information uncertainty affects their use in credit assignment.

## Key findings

- Reduced information uncertainty correlates with increased use of the tabular learning strategy.
- The tabular model outperformed the eligibility trace model under low uncertainty conditions.
- Participants adapted their learning strategies based on environmental uncertainty.

## Abstract

When receiving a reward after a sequence of multiple events, how do we determine which event caused the reward? This problem, known as temporal credit assignment, can be difficult for humans to solve given the temporal uncertainty in the environment. Research to date has attempted to isolate dimensions of delay and reward during decision-making, but algorithmic solutions to temporal learning problems and the effect of uncertainty on learning remain underexplored. To further our understanding, we adapted a reward learning task that creates a temporal credit assignment problem by combining sequentially delayed rewards, intervening events, and varying uncertainty via the amount of information presented during feedback. Using computational modeling, two learning strategies were developed: an eligibility trace, whereby previously selected actions are updated as a function of the temporal sequence, and a tabular update, whereby only systematically related past actions (rather than unrelated intervening events) are updated. We hypothesized that reduced information uncertainty would correlate with increased use of the tabular strategy, given the model’s capacity to incorporate additional feedback information. Both models effectively learned the task, and predicted choices made by participants (N = 142) as well as specific behavioral signatures of credit assignment. Consistent with our hypothesis, the tabular model outperformed the eligibility model under low information uncertainty, as evidenced by more accurate predictions of participants’ behavior and an increase in tabular weight. These findings provide new insights into the mechanisms implemented by humans to solve temporal credit assignment and adapt their strategy in varying environments.

People routinely experience uncertain and temporally extended environments that might be encountered again in the future. To overcome the burden of relearning these environments, people detect patterns that will facilitate future decision-making. Although pattern detection is complex, one such mechanism – known as credit assignment – is particularly important when identifying decisions to repeat and those to avoid. Credit assignment allows people to assign value to past experiences and then utilize these values for quick and efficient decision-making. However, this process becomes substantially more complex when the delay between a decision and its associated outcome increases, and when the available information decreases, making the environment more uncertain. Our research experimentally isolates the effects of delay to understand the strategies people use to solve credit assignment problems in environments with low and high uncertainty. We discover that information uncertainty influences how credit assignment strategies are deployed. Our methods, analyses, and results pave the way for more complex real-world experimentation that seeks to understand the interaction between reward, time, and uncertainty.