arXiv subjects

Michelle Si

Publications and source records attributed to Michelle Si.

3 recordsLinked to original sources

arXiv2026

Algorithmic Impact Reveals the Hidden Social Choice Structure of Alignment

When an AI algorithm makes decisions that affect more than one person, aligning it becomes a problem of social choice: how should people's divergent preferences about system behavior be reconciled and aggregated into a single coherent model? The standard approach to aligning frontier AI models$\unicode{x2013}$reinforcement learning from human feedback$\unicode{x2013}$largely sidesteps this question and has poor social choice guarantees. However, it remains unclear what alternative should replace it. We show that, by focusing directly on an algorithm's welfare consequences, the alignment problem can be reformulated as linear optimization over a convex impact space, which makes it amenable to the standard toolkit of welfare economics and mechanism design. This reformulation clarifies how alignment protocols translate into welfare consequences and, conversely, how a social planner's desired constraints on welfare consequences can be translated back into alignment protocols. We apply this transformation to show that voting-by-issues and random-dictatorship mechanisms are strategyproof and unanimous. Demonstrating the reverse direction, we also apply the impact representation to derive a family of alignment protocols that maximize utilitarian social welfare subject to various social desiderata, such as bounds on individual or group harm. We illustrate the welfare implications of these alignment protocols empirically using real human preferences over kidney allocation, charitable food distribution, LLM responses, and trolley problems.

cs.AI↗

arXiv2026

Internal Pluralism and the Limits of Pairwise Comparisons

Local pairwise comparisons are a standard tool for learning how people want decision rules to work, e.g., in participatory design or alignment. However, their use builds in two strong assumptions: that local comparisons are sufficient evidence about how a person wants an automated decision rule to behave, and that people can always answer those comparisons decisively. We investigate how these assumptions may be compromised under internal pluralism: the idea that an individual evaluates decision rules according to multiple authoritative priorities about how the rule should behave. We provide a formal model of such pluralistic preferences over decision rules, which then lets us identify two distinct failures of forced local pairwise comparison data. First, priorities such as proportionality, egalitarianism, and equal treatment are inherently global: what they imply in one case can depend on what happens elsewhere, so local comparisons may fail to capture them. Second, even when priorities are representable locally, tension between strongly-held priorities can generate internal conflict, producing potentially costly behavioral distortions when comparisons are forced. We then use our model to investigate the alternative -- allowing people to report indecision -- and our findings suggest that doing so can considerably reduce the number of queries needed to learn preferences accurately. We conclude by describing how our model points toward preference-learning methods that elicit these priorities directly, yielding more faithful and interpretable accounts of what people value.

cs.AI↗

arXiv2025

Counterfactual Explanation of Shapley Value in Data Coalitions

The Shapley value is widely used for data valuation in data markets. However, explaining the Shapley value of an owner in a data coalition is an unexplored and challenging task. To tackle this, we formulate the problem of finding the counterfactual explanation of Shapley value in data coalitions. Essentially, given two data owners $A$ and $B$ such that $A$ has a higher Shapley value than $B$, a counterfactual explanation is a smallest subset of data entries in $A$ such that transferring the subset from $A$ to $B$ makes the Shapley value of $A$ less than that of $B$. We show that counterfactual explanations always exist, but finding an exact counterfactual explanation is NP-hard. Using Monte Carlo estimation to approximate counterfactual explanations directly according to the definition is still very costly, since we have to estimate the Shapley values of owners $A$ and $B$ after each possible subset shift. We develop a series of heuristic techniques to speed up computation by estimating differential Shapley values, computing the power of singular data entries, and shifting subsets greedily, culminating in the SV-Exp algorithm. Our experimental results on real datasets clearly demonstrate the efficiency of our method and the effectiveness of counterfactuals in interpreting the Shapley value of an owner.

cs.GT↗