Research map / Bandits and online learning / Multi-armed bandits
Incentivized Bandit Learning with Self-Reinforcing User Preferences
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use three phases: incentivized exploration, incentivized dominance-building, then self-sustaining preference | At-Least-n Explore-Then-Commit · Incentivized bandit with self-reinforcing preferences | expected regret: O(log T) | Helped |
| Use three phases with confidence-based elimination during exploration | UCB-List · Incentivized bandit with self-reinforcing preferences | expected regret: O(log(T)) | Helped |
| Remove incentive control (ablation) | Two-armed simulation with initial bias toward the suboptimal arm · Two-armed simulation, μ = [0.3, 0.5], θ = [100, 1] | average regret | Hurt |
| Use incentives only during exploration, not while building dominance (ablation) | Partial-incentive baseline · Two-armed simulation, μ = [0.3, 0.5], θ = [100, 1] | average regret | Hurt |
| Increase feedback exponent from α = 1.5 to α = 2 (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Three-armed simulation, μ = [0.2, 0.4, 0.6] | average regret | Helped |
| Increase initial biases on suboptimal arms to θ = [50, 50, 1] (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Three-armed simulation, μ = [0.2, 0.4, 0.6] | average regret | Hurt |
| Increase per-step payment from b = 1.2 to b = 1.8 (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Three-armed simulation, μ = [0.2, 0.4, 0.6] | average regret | No clear effect |
| Increase per-step payment from b = 1.2 to b = 1.8 (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Three-armed simulation, μ = [0.2, 0.4, 0.6] | total payment | Hurt |
| Weaken incentive impact to G(b,t) = 0.5 or 0.2 (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Two-armed imperfect-condition simulations | average regret | Hurt |
| Weaken preference feedback to α = 1.0 or 0.2 (ablation) | At-Least-n Explore-Then-Commit and UCB-List · Two-armed imperfect-condition simulations | average regret | Hurt |
| Use UCB-List instead of At-Least-n Explore-Then-Commit | At-Least-n Explore-Then-Commit · Three-armed simulation | average regret | Helped |
Closest papers by meaning
- Bandit Learning with Joint Effect of Incentivized Sampling, Delayed Sampling Feedback, and Self-Reinforcing User Preferences (ICLR 2022 · similarity 0.93)
- Modeling Attrition in Recommender Systems with Departing Bandits (AAAI 2022 · similarity 0.91)
- Bandits Meet Mechanism Design to Combat Clickbait in Online Recommendation (ICLR 2024 · similarity 0.89)
- Bandits with Ranking Feedback (NeurIPS 2024 · similarity 0.89)
- Pairwise Elimination with Instance-Dependent Guarantees for Bandits with Cost Subsidy (ICLR 2025 · similarity 0.88)
- On Multi-Armed Bandit with Impatient Arms (ICML 2024 · similarity 0.88)
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.