Dr. Imara Solveig

Biography

Imara Solveig joined Veyra Institute in 2018 after completing a postdoctoral fellowship at the Halvaard Centre for Responsible Computing, where she developed audit frameworks for black-box algorithmic systems deployed in public-sector decision-making. She received her PhD in human-computer interaction from the University of Telvane in 2015, with a dissertation examining how non-expert users form mental models of opaque recommendation engines. Her undergraduate studies in cognitive science and philosophy were completed at Brennford College, from which she graduated with first-class honours in 2010.

At Veyra, Solveig established the Human-Centred Computing Group to bridge the gap between formal methods in machine learning interpretability and the lived experience of people who are subject to automated decisions. Her group combines qualitative co-design methods with quantitative fairness metrics, producing tools and frameworks that are both theoretically grounded and practically usable by non-specialists. The group's open-source toolkit AuditKit, released in 2022, has been adopted by several public health and financial oversight bodies across the region. She holds a patent on an interactive saliency-based explanation interface and collaborates regularly with Veyra's legal and policy advisory unit on algorithmic accountability standards.

Solveig serves on the programme committee of the Telvane Symposium on Sociotechnical AI and is an elected member of the Institute's Equity and Inclusion Council. She leads the graduate seminar Participatory Methods in AI Research each autumn semester, and contributes to the Institute's executive education programme on responsible technology deployment. She has been principal investigator on three externally funded projects totalling approximately 4.9 million cr, and currently supervises four doctoral students and two postdoctoral researchers.

Research interests

Selected publications

1. Solveig I, Adeyemi R, Kowalska P. "AuditKit: a participatory framework for community-led algorithmic audits." ACM Conference on Fairness, Accountability, and Societal Impact (FACSI), pp. 52–67, 2024. VEYRA-DOI: 10.veyra/VX-2412
2. Solveig I, Mwangi T. "Value conflicts in explainability: when fidelity and comprehensibility diverge." Journal of Human-Centred AI Systems, 6(2): 118–135, 2024. VEYRA-DOI: 10.veyra/VX-2407
3. Kowalska P, Solveig I, Ferrante L. "Saliency maps as co-design artefacts: a field study in clinical decision support." Proceedings of the Velmoor Workshop on Interactive Machine Learning, pp. 31–44, 2023. VEYRA-DOI: 10.veyra/VX-2318
4. Solveig I, Ngozi B. "Disparate impact in ranked recommendation: an intersectional audit protocol." Transactions on Responsible Information Systems, 3(1): 9–28, 2022. VEYRA-DOI: 10.veyra/VX-2209
5. Solveig I, Halvorsen E, Adeyemi R. "Mental models of algorithmic recourse among low-digital-literacy users." Annual Symposium on Human Factors in Automated Systems (HFAS), pp. 200–213, 2021. VEYRA-DOI: 10.veyra/VX-2114
6. Solveig I. "Towards a vocabulary of algorithmic contestability for non-expert stakeholders." Ethics and Information Technology, 23(4): 445–459, 2021. VEYRA-DOI: 10.veyra/VX-2103

Current group members

Postdoctoral researchers

• Dr. Rukayat Adeyemi — participatory audit design, community-centred AI evaluation
• Dr. Piotr Kovalska — interactive saliency methods, user-centred interpretability

Doctoral students

• Tendai Mwangi — value-sensitive design for public-sector AI (Year 3)
• Beatrix Ngozi — intersectional fairness metrics in hiring systems (Year 3)
• Luca Ferrante — contrastive explanation interfaces for clinical AI (Year 2)
• Sigrid Halvorsen — algorithmic recourse under distribution shift (Year 1)