Policy Aggregation

We consider the challenge of AI value alignment with multiple individuals that have different reward functions and optimal policies in an underlying Markov decision process. We formalize this problem as one of policy aggregation, where the goal is to identify a desirable collective policy. We argue that an approach informed by social choice theory is especially suitable. Our key insight is that social choice methods can be reinterpreted by identifying ordinal preferences with volumes of subsets of the state-action occupancy polytope. Building on this insight, we demonstrate that a variety of methods--including approval voting, Borda count, the proportional veto core, and quantile fairness--can be practically applied to policy aggregation.

Honor Among Bandits: No-Regret Learning for Online Fair Division

We consider the problem of online fair division of indivisible goods to players when there are a finite number of types of goods and player values are drawn from distributions with unknown means. Our goal is to maximize social welfare subject to allocating the goods fairly in expectation. When a player's value for an item is unknown at the time of allocation, we show that this problem reduces to a variant of (stochastic) multi-armed bandits, where there exists an arm for each player's value for each type of good. At each time step, we choose a distribution over arms which determines how the next item is allocated. We consider two sets of fairness constraints for this problem: envy-freeness in expectation and proportionality in expectation. Our main result is the design of an explore-then-commit algorithm that achieves $\tilde{O}(T^{2/3})$ regret while maintaining either fairness constraint. This result relies on unique properties fundamental to fair-division constraints that allow faster rates of learning, despite the restricted action space.

Learning Social Welfare Functions

Is it possible to understand or imitate a policy maker's rationale by looking at past decisions they made? We formalize this question as the problem of learning social welfare functions belonging to the well-studied family of power mean functions. We focus on two learning tasks; in the first, the input is vectors of utilities of an action (decision or policy) for individuals in a group and their associated social welfare as judged by a policy maker, whereas in the second, the input is pairwise comparisons between the welfares associated with a given pair of utility vectors. We show that power mean functions are learnable with polynomial sample complexity in both cases, even if the comparisons are social welfare information is noisy. Finally, we design practical algorithms for these tasks and evaluate their performance.

Axioms for AI Alignment from Human Feedback

In the context of reinforcement learning from human feedback (RLHF), the reward function is generally derived from maximum likelihood estimation of a random utility model based on pairwise comparisons made by humans. The problem of learning a reward function is one of preference aggregation that, we argue, largely falls within the scope of social choice theory. From this perspective, we can evaluate different aggregation methods via established axioms, examining whether these methods meet or fail well-known standards. We demonstrate that both the Bradley-Terry-Luce Model and its broad generalizations fail to meet basic axioms. In response, we develop novel rules for learning reward functions with strong axiomatic guarantees. A key innovation from the standpoint of social choice is that our problem has a linear structure, which greatly restricts the space of feasible rules and leads to a new paradigm that we call linear social choice.

Generative Social Choice

Traditionally, social choice theory has only been applicable to choices among a few predetermined alternatives but not to more complex decisions such as collectively selecting a textual statement. We introduce generative social choice, a framework that combines the mathematical rigor of social choice theory with large language models' capability to generate text and extrapolate preferences. This framework divides the design of AI-augmented democratic processes into two components: first, proving that the process satisfies rigorous representation guarantees when given access to oracle queries; second, empirically validating that these queries can be approximately implemented using a large language model. We illustrate this framework by applying it to the problem of generating a slate of statements that is representative of opinions expressed as free-form text, for instance in an online deliberative process.

The Distortion of Binomial Voting Defies Expectation

In computational social choice, the distortion of a voting rule quantifies the degree to which the rule overcomes limited preference information to select a socially desirable outcome. This concept has been investigated extensively, but only through a worst-case lens. Instead, we study the expected distortion of voting rules with respect to an underlying distribution over voter utilities. Our main contribution is the design and analysis of a novel and intuitive rule, binomial voting, which provides strong expected distortion guarantees for all distributions.

The Goal-Gradient Hypothesis in Stack Overflow

According to the goal-gradient hypothesis, people increase their efforts toward a reward as they close in on the reward. This hypothesis has recently been used to explain users' behavior in online communities that use badges as rewards for completing specific activities. In such settings, users exhibit a "steering effect," a dramatic increase in activity as the users approach a badge threshold, thereby following the predictions made by the goal-gradient hypothesis. This paper provides a new probabilistic model of users' behavior, which captures users who exhibit different levels of steering. We apply this model to data from the popular Q&A site, Stack Overflow, and study users who achieve one of the badges available on this platform. Our results show that only a fraction (20%) of all users strongly experience steering, whereas the activity of more than 40% of badge achievers appears not to be affected by the badge. In particular, we find that for some of the population, an increased activity in and around the badge acquisition date may reflect a statistical artifact rather than steering, as was previously thought in prior work. These results are important for system designers who hope to motivate and guide their users towards certain actions. We have highlighted the need for further studies which investigate what motivations drive the non-steered users to contribute to online communities.

Learning and Planning in Feature Deception Games

Today's high-stakes adversarial interactions feature attackers who constantly breach the ever-improving security measures. Deception mitigates the defender's loss by misleading the attacker to make suboptimal decisions. In order to formally reason about deception, we introduce the feature deception game (FDG), a domain-independent game-theoretic model and present a learning and planning framework. We make the following contributions. (1) We show that we can uniformly learn the adversary's preferences using data from a modest number of deception strategies. (2) We propose an approximation algorithm for finding the optimal deception strategy and show that the problem is NP-hard. (3) We perform extensive experiments to empirically validate our methods and results.

Envy-Free Classification

In classic fair division problems such as cake cutting and rent division, envy-freeness requires that each individual (weakly) prefer his allocation to anyone else's. On a conceptual level, we argue that envy-freeness also provides a compelling notion of fairness for classification tasks. Our technical focus is the generalizability of envy-free classification, i.e., understanding whether a classifier that is envy free on a sample would be almost envy free with respect to the underlying distribution with high probability. Our main result establishes that a small sample is sufficient to achieve such guarantees, when the classifier in question is a mixture of deterministic classifiers that belong to a family of low Natarajan dimension.

Choosing How to Choose Papers

It is common to see a handful of reviewers reject a highly novel paper, because they view, say, extensive experiments as far more important than novelty, whereas the community as a whole would have embraced the paper. More generally, the disparate mapping of criteria scores to final recommendations by different reviewers is a major source of inconsistency in peer review. In this paper we present a framework --- based on $L(p,q)$-norm empirical risk minimization --- for learning the community's aggregate mapping. We draw on computational social choice to identify desirable values of $p$ and $q$; specifically, we characterize $p=q=1$ as the only choice that satisfies three natural axiomatic properties. Finally, we implement and apply our approach to reviews from IJCAI 2017.