When I studied visual arts, one lesson that really struck me came from learning to draw. Usually, it can be pretty difficult to draw things realistically. To counter this, we can deliberately try drawing the “negative spaces” — the empty spaces around the object, rather than the object itself (see Figure 1). This simple change in focus can help people draw remarkably more accurately! I use this principle when designing technologies to support thinking. In my research, I explore how technologies can serve as transformative lenses, empowering individuals to adopt new and constructive perspectives. My work investigates this idea across various domains, including creativity, learning, and communication, as showcased in the projects below. Taking a human-centered and research-through-prototyping approach, I uncover key factors and principles that can inform the design of future technologies that empower individuals to purposefully pursue their learning and performance goals.

Designing the visuals for casts of characters is a complex craft, for which there is a lack of tool support. Dedicated avatar and character creation interfaces (e.g., for games and video conferencing), are typically geared towards creating an individual character for a specific universe. General purpose artistic tools, while flexible, do not cater to the entire complex creative workflow. In the process of creating a cast of characters, beyond developing a concept for a single character (e.g., in turnarounds, expression sheets, etc.), designers must fit them within the context of other characters and the larger story (e.g., in lineups and storyboards).

Based on background research and formative interviews with five character designers, we determined that a tool for this task should (1) support rapid instantiations of ideas, (2) allow multi-modal input to maximize expressivity, and (3) enable varied visualizations for contextual thinking.

This led us to create Paratrouper [1], a generative-AI-powered multi-modal authoring tool for visual character cast design (see Figure 2). With it, one can use text, sketches, and image references to create images of original characters within cards. Characters cards can be viewed side-by-side, as well as sorted and styled in groups. Characters can also be visualized from multiple angles in character sheets, and staged together in different settings.

In a user study, we invited eight character designers to use the tool to design an original cast of characters. Based on observations and interviews, we learned that Paratrouper can help creators articulate and refine their design intent through rapid externalization. Character cards encouraged non-linear and parallel design exploration, multi-modal input modalities afforded creative agency and expressivity, and stages fueled holistic thinking as well as sparked new ideas.

[1] Leong, J., Ledo, D., Driscoll, T., Grossman, T., Fitzmaurice, G., & Anderson, F. (2025, April). Paratrouper: Exploratory Creation of Character Cast Visuals Using Generative AI. To appear in Proceedings of the 2025 CHI conference on human factors in computing systems.

[2] Leong, J., Pataranutaporn, P., Danry, V., Perteneder, F., Mao, Y., & Maes, P. (2024, May). Putting things into context: Generative AI-enabled context personalization for vocabulary learning improves learning motivation. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (pp. 1-15).

Public speaking anxiety can get in the way of people sharing their ideas. To deal with this, a common piece of advice is to imagine your audience in a new light. However, what if augmented reality (AR) filters could be used for this instead?

To explore this idea, we conducted a survey (n=100) to capture a snapshot of a general public’s perception of this idea (i.e., acceptance, filter effects, possible motivations, concerns). Furthermore, we prototyped a custom web conferencing application called Masquerade (see Figure 4 and Video 2), that enabled users to apply AR filter effects privately (meaning no one else on the call would see the effects) either on others or on themselves.

In a user study, we invited 16 people with a fear of public speaking (FOPS) to deliver an online speech using the app [3]. We found that AR filters, particularly applied on-others, can be helpful to mitigate public speaking anxiety, via a variety of strategies: moderating the distance they feel to their audience, changing how they feel towards themselves (i.e., more confident, or less self-focused), or altering their perception of the overall situation (i.e., less serious or distracting). Their preferences can depend on the intensity of their anxiety, their existing public speaking habits, and additional personal and social factors. We also uncovered several important ethical considerations. In particular, the ability for all stakeholders to negotiate what features are available, and being transparent about what features are available on each call are critical for maintaining a safe and respectful social atmosphere.

More broadly, the emotions and mindsets we carry govern whether and how we approach experiences that can lead to self growth. In an XRDS magazine article [4], I outlined some of the ways in which I believe generative AI can be used to cultivate positive emotions and mindsets that are conducive to learning and self-development.

[5] Pataranutaporn, P., Danry, V., Leong, J., Punpongsanon, P., Novy, D., Maes, P., & Sra, M. (2021). AI-generated characters for supporting personalized learning and well-being. Nature Machine Intelligence, 3(12), 1013-1022.

[6] Pataranutaporn, P., Leong, J., Danry, V., Lawson, A. P., Maes, P., & Sra, M. (2022, October). AI-generated virtual instructors based on liked or admired people can improve motivation and foster positive emotions for learning. In 2022 IEEE Frontiers in Education Conference (FIE) (pp. 1-9). IEEE.