Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pluralistic Alignment Over Time

Toryn Q. Klassen, Parand A. Alamdari, Sheila A. McIlraith

TL;DR

It is suggested how a recent approach to evaluating fairness over time could be applied to a new form of pluralistic alignment: temporal pluralism, where the AI system reflects different stakeholders' values at different times.

Abstract

If an AI system makes decisions over time, how should we evaluate how aligned it is with a group of stakeholders (who may have conflicting values and preferences)? In this position paper, we advocate for consideration of temporal aspects including stakeholders' changing levels of satisfaction and their possibly temporally extended preferences. We suggest how a recent approach to evaluating fairness over time could be applied to a new form of pluralistic alignment: temporal pluralism, where the AI system reflects different stakeholders' values at different times.

Pluralistic Alignment Over Time

TL;DR

It is suggested how a recent approach to evaluating fairness over time could be applied to a new form of pluralistic alignment: temporal pluralism, where the AI system reflects different stakeholders' values at different times.

Abstract

If an AI system makes decisions over time, how should we evaluate how aligned it is with a group of stakeholders (who may have conflicting values and preferences)? In this position paper, we advocate for consideration of temporal aspects including stakeholders' changing levels of satisfaction and their possibly temporally extended preferences. We suggest how a recent approach to evaluating fairness over time could be applied to a new form of pluralistic alignment: temporal pluralism, where the AI system reflects different stakeholders' values at different times.

Paper Structure

This paper contains 7 sections, 1 equation, 2 figures.

Table of Contents

  1. Temporal Aspects of Pluralistic Alignment
  2. Preference change over time
  3. Temporally extended preferences, norms, goals, and rewards
  4. Pluralistic alignment over time
  5. A framework for pluralistic alignment over time
  6. Notation
  7. Conclusion

Figures (2)

  • Figure 1: Different temporal aspects of pluralistic alignment
  • Figure 2: A reward machine that gives reward 1 only on trajectories on which dessert () is eaten after the main course (). An edge labelled $\langle \varphi, r \rangle$ is taken when the propositional formula $\varphi$ is true, yielding reward $r$.

Theorems & Definitions (2)

  • Definition 1: Temporal pluralism scheme
  • Definition 2: Temporal pluralism score