Delegation as Dialogue: Negotiative Interfaces for Human-AI Conditional Collaboration

Current conditional delegation paradigms (Lai et al., 2022; Spitzer et al., 2024) typically involve humans setting rules or thresholds, with limited two-way communication. Yet, as Hemmer et al. (2024) note, true complementarity arises when both human and AI perspectives are brought to bear on each instance. Drawing inspiration from negotiation theory and collaborative AI research, this idea proposes a new class of interfaces where humans and AI can "discuss" cases: the AI can proactively propose when it believes delegation is warranted, provide reasons (including uncertainty and contextual factors), and accept or challenge human input. Humans, in turn, can ask the AI to justify confidence, flag overlooked factors, or request further information before accepting delegation. Over time, the system could learn negotiation patterns that lead to better outcomes, and even adapt its negotiation strategy to individual users' preferences and expertise. This approach would foster shared agency and mutual learning, moving beyond static delegation toward a dynamic, interactive partnership.

References:

1. Human-AI Collaboration via Conditional Delegation: A Case Study of Content Moderation. Vivian Lai, Samuel Carton, Rajat Bhatnagar, Vera Liao, Yunfeng Zhang, Chenhao Tan, Q. Liao (2022). International Conference on Human Factors in Computing Systems.
2. Complementarity in Human-AI Collaboration: Concept, Sources, and Evidence. Patrick Hemmer, Max Schemmer, Niklas Kuhl, Michael Vossing, G. Satzger (2024). European Journal of Information Systems.
3. Human Delegation Behavior in Human-AI Collaboration: The Effect of Contextual Information. Philipp Spitzer, Joshua Holstein, Patrick Hemmer, Michael Vossing, Niklas Kuhl, Dominik Martin, G. Satzger (2024). Proc. ACM Hum. Comput. Interact..