Arguments of Persuasion in Labour Mediation

TL;DR

This paper introduces PERSUADER, a computational model that generates persuasive arguments for labor mediation by analyzing the goal trees of both the company and the union. It focuses on "threatening" arguments, which highlight how refusing a proposal would violate a party's goals, offering a novel approach to automating conflict resolution in negotiations.

What Problem Does This Paper Address?

Labor negotiations often reach impasses, particularly on secondary issues, requiring a mediator to help parties accept compromises. While argumentation is central to these negotiations, existing research has primarily focused on adversarial arguments or understanding causal relationships in text, not on generating persuasive arguments for cooperative problem-solving. This paper addresses the challenge of how to computationally construct arguments that can effectively persuade parties to accept concessions in a mediation context, specifically by leveraging the parties' underlying goals.

Methods

The core of the PERSUADER system lies in its representation of parties' goals as goal trees. These trees depict hierarchical relationships where higher-level goals are supported by subgoals. The system uses a goal-directed search mechanism to generate arguments. When a mediator needs to convince a party (e.g., the union) about a proposal, the system searches the company's goal tree to identify actions the company might take to offset the union's refusal. These actions must satisfy three conditions: they must be controllable by the company, they must violate a union goal, and the importance of the violated union goal must outweigh the importance of the union's current demand. The process involves propagating the effects of a proposed change (e.g., increased wages) through the goal tree, identifying conflicts, and then searching for compensating actions that threaten the opposing party's goals. The paper also briefly outlines other argument types, such as appeals to universal principles, precedents, and prevailing practices.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Goal Trees as the Foundation for Persuasion: The paper demonstrates that representing parties' goals hierarchically (as trees) is a powerful way to understand their motivations and to derive persuasive arguments.
  • The Power of "Threatening" Arguments: The research highlights the crucial role of "threatening" arguments in labor mediation. By showing how a party's refusal to concede would negatively impact their own important goals, mediators can effectively push for agreement.
  • Computational Approach to Mediation: PERSUADER offers a novel computational framework for assisting mediators by automating the generation of persuasive arguments, moving beyond purely human-driven negotiation strategies.
  • Conflicting Goals as Negotiation Levers: The identification of conflicting goals within a party's own goal tree (e.g., a company wanting to increase economic concessions for efficiency but decrease them for labor costs) reveals internal tensions that can be exploited in negotiation.

Key Results

The paper presents a procedural model for generating threatening arguments. It illustrates the process with two examples:

  • Event 1: A company refuses a wage settlement. The mediator argues that this will lead to employee dissatisfaction, impacting plant efficiency.
  • Event 2: A union refuses a wage settlement. The mediator argues that if the company grants higher wages, it will become noncompetitive and forced to lay off workers.

The paper also walks through a specific example of the PERSUADER program generating an argument for the union regarding wage increases. The system identifies that an increase in wages leads to increased economic costs for the company, which can be compensated by decreasing employment. This action directly threatens the union's goal of employment, which is deemed more important than the wage increase itself, thus forming a persuasive argument.

Limitations and Future Work

The authors acknowledge that their representation of goals is a "crude approximation of reality." They note that the amount by which a quantity is changed is crucial for acceptability, a factor simplified in their model. They assume the mediator has a means of generating reasonable values for demands.

Future work could involve:

  • Generating arguments based on furthering goals (not just threatening them).
  • Developing a more sophisticated understanding of the "convincing power" of arguments, potentially involving a default ranking from weakest to strongest.
  • Incorporating more nuanced representations of goals and their impacts.
  • Addressing the generation of arguments based on furthering goals, which they suggest could be done via a similar search process.

Summary updated

Abstract

In this paper we present a process model of the reasoning underlying arguments of persuasion and its embodiment in a computer program, the PERSUAOER, which gives counsel for the resolution of impasses in collective bargaining. We show how goal trees can be searched to produce arguments involving economic quantities.
BibTeX
@inproceedings{1985arguments,
  title = {{Arguments of Persuasion in Labour Mediation}},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year = {1985},
}

Loading related content...

Figure