Learning to Combine Cues Using Rewards
Author Information
Author(s): Weisswange Thomas H., Rothkopf Constantin A., Rodemann Tobias, Triesch Jochen
Primary Institution: Frankfurt Institute for Advanced Studies, Frankfurt, Germany
Hypothesis
Can reward-dependent learning contribute to the development of cue integration and causal inference?
Conclusion
The study shows that reward-mediated learning can drive the development of cue integration and causal inference.
Supporting Evidence
- Humans combine sensory signals to reduce uncertainty about their causes.
- Bayesian inference models predict human behavior in cue combination tasks.
- Reward-dependent learning is shown to improve cue integration abilities.
Takeaway
This study found that when we learn from rewards, we can get better at combining different types of information, like what we see and hear.
Methodology
The study used a multimodal localization task where learners combined noisy visual and auditory signals to orient towards a target, learning from rewards based on their accuracy.
Limitations
The study does not address how temporal relations between signals affect learning.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website