AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Encoding Human Behavior in Information Design through Deep Learning

Guanghui Yu, Wei Tang, Saumik Narayanan, Chien-Ju Ho

NeurIPS 2023 · Reinforcement learning › Visual and representation RL · approach: Inferring decision-making from behavior · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use HAIDNet policy instead of Bayesian-rational optimal policy (BR-Policy)information policy for human receivers · MTurk phase 2average sender utility: 0.524 -> 0.667Improved
use TH-Model-optimal policy instead of BR-Policyinformation policy · MTurk phase 2average sender utility: 0.524 -> 0.621Improved
replace TH-Model descriptor with data-driven neural network descriptor in policy designHAIDNet · MTurk phase 2average sender utility: 0.621 -> 0.667Improved
replace Bayesian rational descriptor with TH-Model for predicting human choiceshuman behavior descriptor · MTurk phase 1test accuracy: 0.562 -> 0.735Improved
replace TH-Model with neural network descriptorhuman behavior descriptor · MTurk phase 1test accuracy: 0.735 -> 0.770Improved
increase number of states M from 2 to 10HAIDNet · single Bayesian receiver, binary actionsgap to optimal (test utility): 0.7408 vs 0.7451 -> 0.8196 vs 0.8686Hurt
increase number of actions N from 2 to 10HAIDNet · single Bayesian receiver, binary statesgap to optimal (test utility): 0.7408 vs 0.7451 -> 0.6623 vs 0.6963Hurt
increase number of receivers K from 2 to 15HAIDNet · K Bayesian receiversgap to optimal (test utility): 0.7756 vs 0.7873 -> 0.6621 vs 0.6843Mixed
use HAIDNet instead of LP for 18 receiversHAIDNet · run-time, K=18time: 84280 -> 0.352Improved
increase beta_H (more rational receiver)HAIDNet · non-Bayesian receivertest sender utility: 0.5041 -> 0.6964Mixed
train a single HAIDNet for up to 10 receivers instead of fixed KHAIDNet · at most 10 receivers, K=10test sender utility: 0.6650 vs 0.6901Hurt
train a single HAIDNet for up to 5 states/actionsHAIDNet · (M,N)=(2,3)test sender utility: 0.6517 vs 0.7411Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy