Timezone: »

 
Spotlight
Tell me why! Explanations support learning relational and causal structure
Andrew Lampinen · Nicholas Roy · Ishita Dasgupta · Stephanie Chan · Allison Tam · James McClelland · Chen Yan · Adam Santoro · Neil Rabinowitz · Jane Wang · Feilx Hill

Thu Jul 21 11:45 AM -- 11:50 AM (PDT) @ Room 318 - 320

Inferring the abstract relational and causal structure of the world is a major challenge for reinforcement-learning (RL) agents. For humans, language—particularly in the form of explanations—plays a considerable role in overcoming this challenge. Here, we show that language can play a similar role for deep RL agents in complex environments. While agents typically struggle to acquire relational and causal knowledge, augmenting their experience by training them to predict language descriptions and explanations can overcome these limitations. We show that language can help agents learn challenging relational tasks, and examine which aspects of language contribute to its benefits. We then show that explanations can help agents to infer not only relational but also causal structure. Language can shape the way that agents to generalize out-of-distribution from ambiguous, causally-confounded training, and explanations even allow agents to learn to perform experimental interventions to identify causal relationships. Our results suggest that language description and explanation may be powerful tools for improving agent learning and generalization.

Author Information

Andrew Lampinen (DeepMind)

I am interested in cognitive flexibility and generalization, and how these abilities are enabled by factors like language, memory, and embodiment. I am a Senior Research Scientist at DeepMind.

Nicholas Roy (DeepMind)
Ishita Dasgupta (DeepMind)
Stephanie Chan (DeepMind)
Allison Tam (DeepMind)
James McClelland (Stanford University and Deepmind)
James McClelland

I have been using neural networks to model human cognition since the late 1970's, and co-led the <em><Parallel Distributed Processing </em> research group with David Rumelhart in the early 1980's. My main interests lie in capturing human cognitive abilities with neural networks and in using what we know about human cognition to inform the development of better AI systems.

Chen Yan
Adam Santoro (DeepMind)
Neil Rabinowitz (DeepMind)
Jane Wang (DeepMind)
Feilx Hill (Deepmind)

Related Events (a corresponding poster, oral, or spotlight)

More from the Same Authors