# The Best Decisions Are Not the Best Advice: Making Adherence-Aware Recommendations

arXiv: https://arxiv.org/abs/2209.01874 · HTML: https://mlsanity.com/paper/2209.01874/

## Key facts

- Cited by 16 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 55 references that are in the map.
- First posted on arXiv 5 Sep 2022.

## Builds on (most cited first, 7 of 7)

- [European Union regulations on algorithmic decision-making and a "right to explanation"](https://mlsanity.com/paper/1606.08813/) (2016)
- [Non-stationary Stochastic Optimization](https://mlsanity.com/paper/1307.5449/) (2013)
- [Recursive Partitioning for Personalization using Observational Data](https://mlsanity.com/paper/1608.08925/) (2016)
- [Interpretable Optimal Stopping](https://mlsanity.com/paper/1812.07211/) (2018)
- [Lazy-MDPs: Towards Interpretable Reinforcement Learning by Learning When to Act](https://mlsanity.com/paper/2203.08542/) (2022)
- [Robust Policies For Proactive ICU Transfers](https://mlsanity.com/paper/2002.06247/) (2020)
- [Fabrication-Adaptive Optimization, with an Application to Photonic Crystal Design](https://mlsanity.com/paper/1307.5571/) (2013)

## Built on by (most cited first, 16 of 16)

- [Learning to Make Adherence-Aware Advice](https://mlsanity.com/paper/2310.00817/) (2023)
- [AI Agents for Inventory Control: Human-LLM-OR Complementarity](https://mlsanity.com/paper/2602.12631/) (2026)
- [A Q-learning Approach for Adherence-Aware Recommendations](https://mlsanity.com/paper/2309.06519/) (2023)
- [A Framework for Effective AI Recommendations in Cyber-Physical-Human Systems](https://mlsanity.com/paper/2403.05715/) (2024)
- [Expanding Knowledge Graphs with Humans in the Loop](https://mlsanity.com/paper/2212.05189/) (2022)
- [Conformalized Decision Risk Assessment](https://mlsanity.com/paper/2505.13243/) (2025)
- [Narrowing Action Choices with AI Improves Human Sequential Decisions](https://mlsanity.com/paper/2510.16097/) (2025)
- [Balancing Optimality and Diversity: Human-Centered Decision Making through Generative Curation](https://mlsanity.com/paper/2409.11535/) (2024)
- [HR-Bandit: Human-AI Collaborated Linear Recourse Bandit](https://mlsanity.com/paper/2410.14640/) (2024)
- [Human-AI Productivity Paradoxes: Modeling the Interplay of Skill, Effort, and AI Assistance](https://mlsanity.com/paper/2605.11350/) (2026)
- [Reward Design for Justifiable Sequential Decision-Making](https://mlsanity.com/paper/2402.15826/) (2024)
- [The Value of AI Advice: Personalized and Value-Maximizing AI Advisors Are Necessary to Reliably Benefit Experts and Organizations](https://mlsanity.com/paper/2412.19530/) (2024)
- [LIBRA: Language Model Informed Bandit Recourse Algorithm for Personalized Treatment Planning](https://mlsanity.com/paper/2601.11905/) (2026)
- [Strategic Advice in the Age of Personal AI](https://mlsanity.com/paper/2603.02055/) (2026)
- [Power Couple? AI Growth and Renewable Energy Investment](https://mlsanity.com/paper/2603.26678/) (2026)
- [Improving Human Performance with Value-Aware Interventions: A Case Study in Chess](https://mlsanity.com/paper/2604.14465/) (2026)

## Abstract (arXiv)

Many high-stake decisions follow an expert-in-loop structure in that a human operator receives recommendations from an algorithm but is the ultimate decision maker. Hence, the algorithm's recommendation may differ from the actual decision implemented in practice. However, most algorithmic recommendations are obtained by solving an optimization problem that assumes recommendations will be perfectly implemented. We propose an adherence-aware optimization framework to capture the dichotomy between the recommended and the implemented policy and analyze the impact of partial adherence on the optimal recommendation. We show that overlooking the partial adherence phenomenon, as is currently being done by most recommendation engines, can lead to arbitrarily severe performance deterioration, compared with both the current human baseline performance and what is expected by the recommendation algorithm. Our framework also provides useful tools to analyze the structure and to compute optimal recommendation policies that are naturally immune against such human deviations, and are guaranteed to improve upon the baseline policy.