Narrowing Action Choices with AI Improves Human Sequential Decisions
Not on the map. Why?
Abstract (arXiv)
Recent work has shown that, in classification tasks, it is possible to design decision support systems that do not require human experts to understand when to cede agency to a classifier or when to exercise their own agency to achieve complementarityunicodex2014experts using these systems make more accurate predictions than those made by the experts or the classifier alone. The key principle underpinning these systems reduces to adaptively controlling the level of human agency, by design. Can we use the same principle to achieve complementarity in sequential decision making tasks? In this paper, we answer this question affirmatively. We develop a decision support system that uses a pre-trained AI agent to narrow down the set of actions a human can take to a subset, and then asks the human to take an action from this action set. Along the way, we also introduce a bandit algorithm that leverages the smoothness properties of the action sets provided by our system to efficiently optimize the level of human agency. To evaluate our decision support system, we conduct a large-scale human subject study (n = 1,600) where participants play a wildfire mitigation game. We find that participants who play the game supported by our system outperform those who play on their own by ~30% and the AI agent used by our system by >2%, even though the AI agent largely outperforms participants playing without support. We have made available the data gathered in our human subject study as well as an open source implementation of our system at https://github.com/Networks-Learning/narrowing-action-choices .
Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 34 of its 50 references that are in the map. First posted on arXiv 17 Oct 2025.
Builds on
34 of 50 references.
References · 34
- 2013Playing Atari with Deep Reinforcement Learningcited by 3,211
- 2015Deep Reinforcement Learning with Double Q-learningcited by 1,704
- 2019Dota 2 with Large Scale Deep Reinforcement Learningcited by 1,084
- 2019Introduction to Multi-Armed Banditscited by 475
- 2020Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performancecited by 396
- 2020Effect of Confidence and Explanation on Accuracy and Trust Calibration in AI-Assisted Decision Makingcited by 281
- 2017StarCraft II: A New Challenge for Reinforcement Learningcited by 268
- 2020Consistent Estimators for Learning to Defer to an Expertcited by 262
- 2022Pre-Trained Language Models for Interactive Decision-Makingcited by 220
- 2020A Closer Look at Invalid Action Masking in Policy Gradient Algorithmscited by 160
- 2019Mastering Complex Control in MOBA Games with Deep Reinforcement Learningcited by 150
- 2019Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Modelscited by 114
22 more
- 2019The Algorithmic Automation Problem: Prediction, Triage, and Human Effortcited by 101
- 2021Towards a Science of Human-AI Decision Making: A Survey of Empirical Studiescited by 85
- 2021Differentiable Learning Under Triagecited by 74
- 2022Improving Expert Predictions with Conformal Predictioncited by 54
- 2020Classification Under Human Assistancecited by 42
- 2019Regression Under Human Assistancecited by 40
- 2021Combining Human Predictions with Model Probabilities via Confusion Matrices and Calibrationcited by 40
- 2021Counterfactual Explanations in Sequential Decision Making Under Uncertaintycited by 36
- 2023Human-Aligned Calibration for AI-Assisted Decision Makingcited by 28
- 2023Designing Decision Support Systems Using Counterfactual Prediction Setscited by 26
- 2020Misplaced Trust: Measuring the Interference of Machine Learning in Human Decision-Makingcited by 25
- 2024Towards Human-AI Complementarity with Prediction Setscited by 25
- 2022The Best Decisions Are Not the Best Advice: Making Adherence-Aware Recommendationscited by 16
- 2021Reinforcement Learning Under Algorithmic Triagecited by 15
- 2021Improving Human Sequential Decision-Making with Reinforcement Learningcited by 12
- 2023Auditing for Human Expertisecited by 12
- 2023A Lipschitz Bandits Approach for Continuous Hyperparameter Optimizationcited by 9
- 2025Towards Human-AI Complementarity in Matching Taskscited by 7
- 2022Safe and Psychologically Pleasant Traffic Signal Control with Reinforcement Learning using Action Maskingcited by 6
- 2019Preselection Banditscited by 5
- 2025Human-Alignment Influences the Utility of AI-assisted Decision Makingcited by 5
- 2020Learning to Switch Among Agents in a Team via 2-Layer Markov Decision Processescited by 3
Built on by
Citing papers per year: 2026: 3.
3 papers.