Integrating Fairness in the Software Design Process: An Interview Study With HCI and ML Experts

The term Fairness is used in the field of Machine Learning to refer a suite evaluation techniques, measures, and process adjustments focused on reduction bias statistical models. With many different potential definitions methodologies for implementing fairness, some contradictory goal or incompatible approach, there are challenges selection most appropriate definition(s) meet ethical expectations society end users may have model. This further confounded by potentially complex engineering applying selected definition. designing developing fair system requires interdisciplinary collaboration between user experience researchers, socio-technical application usage adoption, data scientists machine learning software developers, directly involved technical optimisation such How these two groups interact how their areas specialisation can support influence other when AI unknown. leaves open questions about stakeholders view it be operationalised current work, what obstacles/opportunities improving treatment fairness practice. In order answer questions, 18 subject matter experts from fields human-computer interaction (HCI) (ML) were interviewed with perception they would anticipate set example scenarios. work identifies artificial intelligence (AI). dealing challenges, we propose approaches that include expanding responsibilities HCI ML specialists, identifying where joint responsibility required.