Source-linked AI summary
Bridging Formal and Perceived Fairness: Development of an Interdisciplinary Framework in Algorithmic Decision-Making
Maike Lindermayr, Mattia Cerrato, Luisa Hübner, Johannes Kraus
TL;DR
Algorithmic fairness research lacks an integrated account of how formal fairness criteria relate to perceived fairness in sociotechnical settings. This work develops a conceptual framework through literature synthesis, interdisciplinary workshops, and stakeholder interviews, aiming to connect computational audits with user-centered assessments and human-centered design. The framework is currently an early-stage proposal intended to support informed, well-calibrated fairness judgments.
Problem
Algorithmic fairness research is predominantly shaped by Computer Science and formal metrics, while fairness perceptions also depend on cognitive, normative, experiential, and sociotechnical factors.
Method
The project combines theoretical literature synthesis, interdisciplinary workshops, and qualitative semistructured stakeholder interviews to develop and refine an interdisciplinary framework.
Results
The project presents a work-in-progress conceptual framework linking formal fairness properties, sociotechnical mediators, mental models, perceived fairness, trust, and acceptance.
Takeaways & Limitations
The framework aims to support hybrid fairness evaluations and human-centered design that address both technical criteria and how affected stakeholders judge fairness.
Abstract
from arXiv · showhide
While fairness has become a central concern in research on algorithmic systems, the field remains predominantly shaped by Computer Science, resulting in a strong emphasis on formal fairness metrics and bias mitigation strategies. Nevertheless, this focus may obscure a fundamental challenge: fairness is not merely a technical property, but a subjective, context-sensitive human judgment shaped by cognitive heuristics, mental models, normative expectations, and sociotechnical factors. Crucially, users' perceptions of fairness may diverge substantially from the fairness criteria an algorithm formally satisfies; a system may meet predefined technical fairness requirements yet still be perceived as unjust by decision-affected stakeholders. In such cases, the system fails on a fundamental dimension: it will not be trusted, accepted, or considered legitimate. Taking a user-centered design perspective, this paper presents a work-in-progress conceptual framework that bridges Computer Science approaches to formal algorithmic fairness with normative and Social Science fairness approaches regarding perceived fairness, trust, and technology acceptance, embedding both within the sociotechnical conditions that shape human judgment. Through (1) theoretical literature synthesis, (2) interdisciplinary workshops, and (3) stakeholder interviews, the project aims to inform evaluation approaches that integrate computational fairness audits with user-centered assessments and guide the design of fairness-aware, human-centered algorithmic systems that support informed, well-calibrated fairness judgments by those affected.
1 Introduction and Background
Algorithmic fairness research is dominated by formal, auditable metrics, but fairness also depends on human judgments shaped by individual, organizational, and societal context. Formal criteria and perceived fairness are complementary, yet an integrated sociotechnical framework remains lacking.
- 1 Introduction and Background: Formal criteria such as demographic parity and equalized odds define fairness as measurable properties of algorithmic outputs.These criteria support precise auditing and regulatory accountability.
- 1 Introduction and Background: Fairness judgments are shaped by users’ values, experiences, cognitive processing, organizational constraints, domain norms, and societal context.The paper identifies mental models—beliefs about a system’s contents, operation, and rationale—as central to perceived fairness.
- 1 Introduction and Background: Formal metrics may align with distributive justice while underspecifying procedural and interactional concerns such as explanation, respectful treatment, and contestability.This creates conceptual divergences between technical fairness and broader justice judgments.
- 1 Introduction and Background: Formal criteria provide auditability and accountability, whereas perceived fairness influences whether systems are trusted, accepted, and used.The paper argues that both perspectives are necessary for systems that support informed, well-calibrated fairness judgments.
- 1 Introduction and Background: Research lacks an integrated framework linking formal and perceived fairness while accounting for the sociotechnical conditions under which perceptions emerge.The paper addresses this gap through a user-centered design lens.
2 Expected Contributions
The proposed framework integrates formal fairness criteria with normative and social-science accounts of fairness perception. It is intended to support hybrid evaluation and human-centered design that treats perceived legitimacy as a first-order concern.
- 2 Expected Contributions: The framework maps why technical fairness requirements do not necessarily produce perceived fairness by incorporating experiential, normative, relational, and sociotechnical factors.It provides an interdisciplinary structure for research linking Computer Science, Human Factors, Social Science, and Law.
- 2 Expected Contributions: The framework proposes hybrid evaluations combining computational audits of formal criteria with structured user-centered assessments of perceived fairness.This approach aims to reduce deployment failures arising from perceived unfairness.
- 2 Expected Contributions: Design guidance is intended to address how algorithmic decisions are communicated, contextualized, explained, and made contestable so affected people perceive them as legitimate.The proposal makes human experience of fairness a first-order design criterion rather than an afterthought.
3 Methodological Approach
The project develops its interdisciplinary framework through literature synthesis, cross-disciplinary workshops, and stakeholder interviews. These stages map conceptual relationships and refine the framework using stakeholder mental models and normative judgments.
- 3 Methodological Approach: The literature synthesis maps formal fairness metrics against psychological, justice, sociotechnical, trust, acceptance, and behavioral-intention constructs.It identifies overlaps, tensions, and trade-offs that formal criteria cannot capture, producing a preliminary construct taxonomy.
- 3 Methodological Approach: Structured workshops establish shared definitions, identify mediators of perceived fairness, trust, and acceptance, and negotiate integration across disciplines.Their anticipated output specifies direct effects, mediators, and moderators linking formal properties, sociotechnical factors, and user outcomes.
- 3 Methodological Approach: Semistructured interviews with diverse stakeholders examine mental models, legitimate decision attributes, group differences, prior experience, transparency, and societal values.Findings will refine the framework, identify unmodeled variables, and support user-centered design guidelines.
4 Current Status and Future Directions
The project is in an early theoretical phase, with an initial framework describing how system information may shape mental models and fairness judgments. Future work will validate the framework quantitatively while modeling sociotechnical moderation and reciprocal evaluation.
- 4 Current Status and Future Directions: The current framework treats fairness-related system properties and responsible actors as sources of information that users process through central or peripheral routes.This processing contributes to forming a mental model of AI-based decision-making.
- 4 Current Status and Future Directions: Sociotechnical moderators such as time pressure, motivation, and AI literacy may alter cognitive processing and reliance on heuristics.The framework therefore treats fairness perception as context-dependent rather than isolated from user conditions.
- 4 Current Status and Future Directions: Acceptance and other system evaluations may reinforce or revise fairness judgments over time, so the proposed relationships are potentially reciprocal.The framework does not assume that downstream acceptance is determined by fairness perceptions alone.
- 4 Current Status and Future Directions: Subsequent phases will pursue quantitative validation of the initial conceptual framework.The project invites feedback to refine the framework and identify directions for integrating technical and human-centered perspectives.