Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Echoes of AI Harms: A Human-LLM Synergistic Framework for Bias-Driven Harm Anticipation

Nicoleta Tantalaki, Sophia Vei, Athena Vakali

TL;DR

ECHO introduces a proactive, human-centered framework that links AI bias types to potential harms across sociotechnical contexts. By leveraging domain-specific vignettes, dual human-LLM harm annotation, and ethical matrices, it maps bias origins through the AI lifecycle to stakeholder-specific harms in two high-stakes domains (disease diagnosis and hiring). The paper provides descriptive and inferential ethical matrices (dEM and iEM) to reveal robust bias–harm pathways and demonstrates how these insights can guide early design and governance decisions. This approach advances anticipatory governance in AI, offering a generalizable protocol for tracing harm pathways from biases to outcomes prior to deployment or development, with clear implications for policy and risk management.

Abstract

The growing influence of Artificial Intelligence (AI) systems on decision-making in critical domains has exposed their potential to cause significant harms, often rooted in biases embedded across the AI lifecycle. While existing frameworks and taxonomies document bias or harms in isolation, they rarely establish systematic links between specific bias types and the harms they cause, particularly within real-world sociotechnical contexts. Technical fixes proposed to address AI biases are ill-equipped to address them and are typically applied after a system has been developed or deployed, offering limited preventive value. We propose ECHO, a novel framework for proactive AI harm anticipation through the systematic mapping of AI bias types to harm outcomes across diverse stakeholder and domain contexts. ECHO follows a modular workflow encompassing stakeholder identification, vignette-based presentation of biased AI systems, and dual (human-LLM) harm annotation, integrated within ethical matrices for structured interpretation. This human-centered approach enables early-stage detection of bias-to-harm pathways, guiding AI design and governance decisions from the outset. We validate ECHO in two high-stakes domains (disease diagnosis and hiring), revealing domain-specific, bias-to-harm patterns and demonstrating ECHO's potential to support anticipatory governance of AI systems

Echoes of AI Harms: A Human-LLM Synergistic Framework for Bias-Driven Harm Anticipation

TL;DR

ECHO introduces a proactive, human-centered framework that links AI bias types to potential harms across sociotechnical contexts. By leveraging domain-specific vignettes, dual human-LLM harm annotation, and ethical matrices, it maps bias origins through the AI lifecycle to stakeholder-specific harms in two high-stakes domains (disease diagnosis and hiring). The paper provides descriptive and inferential ethical matrices (dEM and iEM) to reveal robust bias–harm pathways and demonstrates how these insights can guide early design and governance decisions. This approach advances anticipatory governance in AI, offering a generalizable protocol for tracing harm pathways from biases to outcomes prior to deployment or development, with clear implications for policy and risk management.

Abstract

The growing influence of Artificial Intelligence (AI) systems on decision-making in critical domains has exposed their potential to cause significant harms, often rooted in biases embedded across the AI lifecycle. While existing frameworks and taxonomies document bias or harms in isolation, they rarely establish systematic links between specific bias types and the harms they cause, particularly within real-world sociotechnical contexts. Technical fixes proposed to address AI biases are ill-equipped to address them and are typically applied after a system has been developed or deployed, offering limited preventive value. We propose ECHO, a novel framework for proactive AI harm anticipation through the systematic mapping of AI bias types to harm outcomes across diverse stakeholder and domain contexts. ECHO follows a modular workflow encompassing stakeholder identification, vignette-based presentation of biased AI systems, and dual (human-LLM) harm annotation, integrated within ethical matrices for structured interpretation. This human-centered approach enables early-stage detection of bias-to-harm pathways, guiding AI design and governance decisions from the outset. We validate ECHO in two high-stakes domains (disease diagnosis and hiring), revealing domain-specific, bias-to-harm patterns and demonstrating ECHO's potential to support anticipatory governance of AI systems

Paper Structure

This paper contains 32 sections, 10 equations, 5 figures, 9 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related work
  3. AI Harms taxonomies
  4. AI bias in a project's lifecycle: Types and mitigation pathways
  5. Human-Centered analytical structures: Ethical Matrices and Vignettes
  6. ECHO: Preliminaries and Framework Overview
  7. Notation and Core Definitions
  8. ECHO Algorithm and Workflow
  9. ECHO in practice: Experimental Setup
  10. Protocol
  11. Participants
  12. An Illustrative ECHO Instance
  13. Results and observations
  14. Descriptive Ethical Matrix
  15. Descriptive Visualization of Bias--Harm Patterns Across Stakeholders and Domains
  16. ...and 17 more sections

Figures (5)

  • Figure 1: Overview of the ECHO workflow. 1. The system takes as input an application domain and generates a domain-specific list of stakeholders. 2. It then takes this list, a list of bias, and a list of harms. For each (domain, bias) pair, it constructs a vignette that is combined with every stakeholder in the domain. Each resulting vignette is then augmented with a MCQ based on the list of harms. 3. The resulting augmented vignette is then annotated by multiple human annotators and an AI agent, each selecting potential harms. The collected annotations are aggregated into a descriptive ethical matrix, where rows represent stakeholders, columns represent bias types, and each cell reflects the consensus view of perceived harms. This matrix is finally refined via inferential testing yielding the inferential ethical matrix.
  • Figure 2: Bias x harm radar plots for (a) Patients and (b) Applicants.
  • Figure 3: Bias × representational harm radar plots for (a) the Patient group and (b) the Applicant group.
  • Figure 4: Bias × harm radar plots for (a) the developer of the Disease Diagnosis System and (b) the developer of the Hiring System.
  • Figure 5: Bias × harm radar plots for (a) the Healthcare Institution and (b) the Company.

Theorems & Definitions (5)

  • definition 1
  • definition 2
  • definition 3
  • definition 4
  • definition 5