Internal Pluralism and the Limits of Pairwise Comparisons

内部多元主义与成对比较的局限性

Abstract: 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.

摘要： 局部成对比较（Local pairwise comparisons）是了解人们期望决策规则如何运作的标准工具，例如在参与式设计或人工智能对齐中。然而，这种方法建立在两个强假设之上：一是局部比较足以作为证据，反映个人希望自动化决策规则如何表现；二是人们总是能够果断地回答这些比较问题。

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.

我们研究了在“内部多元主义”（Internal Pluralism）下，这些假设可能如何受到损害。内部多元主义是指个人根据关于规则应如何表现的多种权威优先事项来评估决策规则的观点。我们为这种针对决策规则的多元偏好提供了一个形式化模型，并据此识别出强制性局部成对比较数据存在的两种截然不同的缺陷。

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.

首先，诸如比例性、平等主义和平等待遇等优先事项本质上是全局性的：它们在某种情况下的含义可能取决于其他地方发生的事情，因此局部比较可能无法捕捉到这些特征。其次，即使优先事项可以在局部表示，强烈持有的优先事项之间的张力也可能产生内部冲突，当被迫进行比较时，会导致潜在的代价高昂的行为扭曲。

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.

随后，我们利用该模型研究了另一种选择——允许人们表达“不确定性”。研究结果表明，这样做可以显著减少准确学习偏好所需的查询次数。最后，我们描述了该模型如何指向那些直接引出这些优先事项的偏好学习方法，从而对人们的价值观得出更忠实、更具可解释性的描述。