Search

Summary

The ability to seek out new information is crucial in many situations of our everyday lives. In general, people can display quite elaborated exploration behavior. However, exploration has mainly been studied in multiarmed bandit tasks and theories have predominantly focused on simple directed and random exploration strategies. In this chapter, we review the results of prior studies and argue that the repertoire of human exploration strategies is much more diverse than how it is portrayed in the literature. However, to find evidence for more sophisticated strategies, more complex paradigms than multiarmed bandits are required. In particular, we argue that Markov Decision Processes offer an interesting new setting that allows us to capture strategies beyond random and directed exploration, such as empowerment-based strategies or strategies that explore using explicit goals. We conclude this chapter by discussing several new experimental paradigms that could advance our understanding of human exploration to the next level.

We give a very general reformulation of multi-actor Markov decision processes and show that there is a tendency for the actors to take the same action whenever possible. This considerably reduces the complexity of the problem, either facilitating numerical computation of the optimal policy or providing a basis for a heuristic.

Search Results

Refine search

Refine search

Actions for selected content:

2 results

Chapter 7 - Exploration Beyond Bandits

Summary

Multi-Actor Markov Decision Processes

Search Results

Refine search

Refine search

Actions for selected content:

Save Search

2 results

Chapter 7 - Exploration Beyond Bandits

Summary

Multi-Actor Markov Decision Processes