AI for All: Identifying AI incidents Related to Diversity and Inclusion

Authors: Rifat Ara Shams, Didar Zowghi, Muneera Bano

JAIR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental This study addresses these challenges by identifying and understanding D&I issues within AI systems through a manual analysis of two AI incident databases, AI Incident Database (AIID) and AI, Algorithmic, and Automation Incidents and Controversies (AIAAIC). The research develops a decision tree to investigate D&I issues tied to AI incidents and populate a public repository of D&I-related AI incidents. The decision tree was validated through a card sorting exercise and focus group discussions. The research demonstrates that almost half of the analyzed AI incidents are related to D&I, with a notable predominance of racial, gender, and age discrimination.
Researcher Affiliation Academia RIFAT ARA SHAMS , CSIRO s Data61, Australia DIDAR ZOWGHI, CSIRO s Data61, Australia MUNEERA BANO, CSIRO s Data61, Australia
Pseudocode Yes Figure 3 shows the results of RQ1. To identify Diversity and Inclusion (D&I) related AI incidents, we developed a decision tree after a rigorous analysis. Our proposed decision tree has four conditions. The first condition checks if any human is directly or indirectly impacted by the AI incident. ... Similarly, Figure 4 (b) shows an example of an AI incident that is not related to D&I issues.
Open Source Code No The paper does not provide explicit source code for its methodology (e.g., the implementation of the decision tree or the analysis scripts). It does mention a public repository of D&I-related AI incidents, but this is a data resource, not source code for their method: "Finally, we have designed, populated, and made publicly available repository of D&I-related AI incidents, intended for use for future research." and "Rifat Ara Shams, Didar Zowghi, and Muneera Bano. 2024. Diversity and Inclusion (DI)Related AI Incidents Repository. https://doi.org/10.5281/zenodo.11639709"
Open Datasets Yes This study addresses these challenges by identifying and understanding D&I issues within AI systems through a manual analysis of two AI incident databases, AI Incident Database (AIID) and AI, Algorithmic, and Automation Incidents and Controversies (AIAAIC). ... 1https://incidentdatabase.ai/ 2https://www.aiaaic.org/aiaaic-repository ... Finally, we created a public repository of D&I-related AI incidents for both AIID and AIAAIC. This open-source repository will serve as a valuable resource for other researchers in this field. ... Rifat Ara Shams, Didar Zowghi, and Muneera Bano. 2024. Diversity and Inclusion (DI)Related AI Incidents Repository. https://doi.org/10.5281/zenodo.11639709
Dataset Splits No The paper describes a qualitative and mixed-methods study involving manual analysis of incidents, card sorting, and focus group discussions. It does not perform machine learning experiments that would typically involve training, validation, and test dataset splits.
Hardware Specification No The paper describes a qualitative study involving manual analysis, card sorting exercises, and focus group discussions. It does not mention any computational experiments requiring specific hardware specifications.
Software Dependencies No The paper describes a qualitative and mixed-methods study (manual analysis, card sorting, focus groups). It does not mention specific software dependencies with version numbers for computational experiments.
Experiment Setup No The paper describes a qualitative study with a focus on identifying and categorizing AI incidents, developing a decision tree, and validating it through participatory activities. It does not involve computational experiments with hyperparameters or system-level training settings.