A brief summary of INTERACT 2023
Last week (29–31 Aug), I attended INTERACT 2023 — one of the longest running conferences on human-computer interaction. Despite the unexpected train strikes, I managed to attend two full days of the main conference. In this blog post, I’ll provide a brief summary of some of the standout works that captured my attention. For those eager to dive deeper, the full program is available online.
Under the theme of “Design for Equality and Justice,” the program featured a diverse range of topics. In what follows, I’ll focus on selected sessions I attended, covering user studies, fairness, explainable and social AI.
On the first day, I presented our work on exploring the link between personality traits and mobile engagement strategies. User engagement plays a critical role in the success of mobile apps. However, most apps follow a one-size-fits-all approach, where all users are exposed to the same engagement strategies. To test whether the very same techniques may be perceived differently by different individuals, we developed and deployed a location-based Augmented Reality (AR) mobile app called HarrySpotter (inspired by the popular Harry Potter series authored by J.K. Rowling), which allows users to annotate real-world objects. Over the course of a two-week in-the-wild study, 29 participants engaged with HarrySpotter, resulting in a collection of 503 annotated objects, and self-reported their personality using the short 10-item TIPI questionnaire. By analyzing the engagement techniques chosen by the participants, we found that correlations between these techniques and users’ Big Five personality traits. In particular, we found that:
- techniques fostering competition were found to discourage agreeable users while encouraging conscientious users.
- techniques promoting exploration of objects and places appealed to extroverts and individuals open to new experiences.
- techniques focusing on personal achievements were found to motivate neurotic users.
Understanding the interplay between users’ personality traits and engagement techniques allows app developers to unlock new possibilities for enhancing user satisfaction and retention. However, it is also important to recognize that this capability cannot only enrich user satisfaction and retention but also raise a multitude of ethical questions. The historical misuse of personality predictions, exemplified by instances where they were used to infer political orientation or disseminate misinformation, underscores the need for their responsible use.
My presentation took place in the “User Studies” session, where two other papers explored the effect of nature on well-being and stress reduction. In particular, one paper highlighted the potential of fractal-based geometries in virtual reality (VR) environments for restoration purposes. In the another session, the importance of mitigating human cognitive biases early in the data entry generation process was emphasized. The authors developed a conversational agent with voice-based data entry and visualization to support financial consultations, aiming to reduce biases caused by inaccurate data and forgetfulness during note-taking. This reminded me Michael Muller and Angelika Strohmayer’s work on “forgetting practices in data sciences.” Another thought-provoking paper questioned whether the study design itself could contribute to overreliance on AI. By comparing various study designs, the authors found indications that certain designs might inadvertently encourage overreliance. While switching rooms, I bumped into a poster, titled “Are Italian and French public university websites sustainable?”. The short answer is that Italian universities are more sustainable than the French ones, but both are far from being sustainable — sustainability is a huge topic, and more needs to be done.
The second day featured a series of interesting papers in the social AI session. One paper, titled “AI in the Human Loop: The Impact of Differences in Digital Assistant Roles on the Personal Values of Users,” urged us to consider the social roles of AI systems in shaping meaningful human-AI interactions beyond their functional capabilities. Another study explored individuals’ expectations regarding the use of smart streetlights, highlighting that participants were more accepting of data collection when it benefited the community and occurred through non-intrusive, privacy-preserving sensors. These findings offer valuable insights for resident-centered smart streetlight design, urban planning, and policymaking. The day concluded with a panel discussion titled “A Multi-Perspective Panel on User-Centered Transparency, Explainability, and Controllability in Automations.” Esteemed researchers, including Philippe Palanque, Fabio Paterno, Virpi Roto, Albrecht Schmidt, Simone Stumpf, and Jürgen Ziegler, shared their perspectives. Simone emphasized the need to shift our focus from explainability to intelligibility, highlighting that it’s essential for users to build appropriate mental models when interacting with AI systems. Albrecht delved into the amplification of the human mind through generative AI, arguing that there will be a fundamental transformation of our knowledge and information culture — a transformation that will be even more radical and much quicker than the the shift from oral traditions to societies based on writing.
Overall, my experience at INTERACT 2023 in York was nothing short of remarkable. York is a beautiful city with lots of history. Now, looking forward to INTERACT 2025, which is set to take place in Brazil — an exciting prospect for fostering more conferences and collaborations in Latin America.
