PhD Project by: Malak Sadek (mfs21@ic.ac.uk)
Supervised by: Dr. Celine Mougenot, Dr. Rafael Calvo

To learn more about the project please email Malak Sadek or follow the project’s page on Medium

Conversational agents (CAs) such as chatbots, smart home speakers, and voice assistants are infiltrating an increasing number of critical applications and domains such as healthcare, education, and finance. The recent proliferation of Generative AI and Large Language Models has only sped up the development and adoption of this technology. With this increasing usage comes increasing concerns over the socio-technical implications and the impacts and consequences of CAs on various groups of stakeholders, especially given their technical nature and opacity. Several recommendations have been made recently for more diverse stakeholder participation in the creation of CAs to make them more human-centred.

As such, this project looks at how collaborative design can be leveraged to better align CAs with human values. Using Value-Sensitive Design as a scaffolding, the result is a framework and a corresponding toolkit that can allow teams building CAs to do so more collaboratively with different stakeholders in ways that better embody their values. It offers a practical step towards creating more value-sensitive CAs.

The project follows a Design Research Methodology format. The first phase of the project consisted of a literature review and a series of semi-structured interviews with CA practitioners. The aim was to understand (i) the limitations of current design interventions and (ii) the challenges and barriers faced by practitioners in terms of collaboratively creating value-sensitive CAs. This phase helped establish the need for a value-sensitive framework and derive a number of design requirements and evaluation metrics for the proposed framework. The second phase involved developing the framework and the toolkit, as well as creating an evaluation protocol for validating them during the final phase.

The final phase is now underway and focuses on evaluating the framework and toolkit. Design workshops with industry practitioners are taking place where they use the framework and toolkit in a fictional case study to prototype a CA. The planned mixed-method evaluation includes semi-structued interviews with workshop participants revolving around the effectiveness of value elicitation and the technical utility of outcomes produced, and a quantitative survey comparing the value alignment of the framework-enabled prototypes with a control prototype created through collaborative design, but without the framework. The findings of this study will (i) evaluate the framework and toolkit and assess their impact, and (ii) provide insights into practically supporting the collaborative design of value-sensitive CAs and AI in general.

Publications

Sadek, M., Constantinides, M., Quercia, D. and Mougenot, C. (2024). Guidelines for Integrating Value Sensitive Design in Responsible AI Toolkits. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (CHI’24). Association for Computing Machinery, New York, NY, USA. https://doi.org/10.1145/3613904.3642810

Sadek, M., Kallina, E., Bohné, T., Mougenot, C., Calvo, R.A., Cave, S. (2024). Challenges of responsible AI in practice: scoping review and recommended actions. AI & Society. https://doi.org/10.1007/s00146-024-01880-9

Sadek, M., Calvo, R.A., Mougenot, C. (2023). Co-designing conversational agents: A comprehensive review and recommendations for best practices, Design Studies, 89. https://www.sciencedirect.com/science/article/pii/S0142694X23000716

Sadek, M., Calvo, R.A., Mougenot, C. (2023). Designing value‐sensitive AI: a critical review and recommendations for socio‐technical design processes, AI Ethics, DOI: https://doi.org/10.1007/s43681-023-00373-7

Sadek, M., Calvo, R.A., Mougenot, C. (2023). Trends, Challenges and Processes in Conversational Agent Design: Exploring Practitioners’ Views through Semi-Structured Interviews. In Proceedings of the 5th International Conference on Conversational User Interfaces (CUI’23). Association for Computing Machinery, New York, NY, USA, Article 13, 1-10. https://doi.org/10.1145/3571884.3597143

Sadek, M., Calvo, R.A., Mougenot, C. (2023). The Value-Sensitive Conversational Agent Co-Design Framework. ArXiV pre-print. https://doi.org/10.48550/arXiv.2310.11848

Sadek, M., Mougenot, C. (2023). Exploring Links between Conversational Agent Design Challenges and Interdisciplinary Collaboration. ArXiV pre-print. https://doi.org/10.48550/arXiv.2311.08832

Sadek, M., Calvo, R.A., Mougenot, C. (2023). Why codesigning AI is different and difficult. ACM Interactions. https://interactions.acm.org/blog/view/why-codesigning-ai-is-different-and-difficult

About Malak

Malak is a doctoral candidate at the Dyson School of Design Engineering at Imperial College London and an Enrichment Student at The Alan Turing Institute. She also holds fellowships at The Leverhulme Centre for the Future of Technology and the Centre for Human-Inspired Artificial Intelligence at Cambridge University.

Malak earned a Bachelor's degree in Computer Engineering with a minor in Digital Media from the American University in Cairo, Egypt (Summa Cum Laude - Highest Honours); and a Master's degree in Human-Computer Interaction from the University of St Andrews, Scotland (Distinction).