HCI news delivered in chronological order.
Get the weekly HCI highlights delivered to your inbox every Friday.
Q. As agents take on more work, why does the problem shift from ‘can they do it’ to ‘can we trust them enough to hand it over’?
A. Because simply performing well isn’t enough. In real work, it’s more important whether the result...
Q. I understood that the core of this study isn’t simply about displaying heart rate—it’s about creating a loose yet real connection between strangers. But why did you feel that a bodily experience like vibration and trembling fit better than numbers or a screen?
A. Experiencing it through the body makes people receive the data more like “energy” than “informati...
Q. Why do you think explainable ML is more important than simply a ‘feature that shows results’?
A. Explanations are not decoration that makes results look nicer; they are a mechanism that lets cli...
Q. Why is it important to look at multiple signals together—can’t you just rely on one thing like voice or facial expressions?
A. Relying on a single signal makes it highly likely to misread the situation. For example, a person...
Q. In this study, what do you think was the biggest reason the EMA platform increased adolescent participation? I’m curious about what made adolescents feel like the research was ‘their own’—not just that the app was convenient.
A. The biggest reason was that the participation process fit adolescents’ everyday lives well. They ...
Q. In this article, the author suggests that matching people’s behavior and producing good outcomes for people are different things. Why does this gap keep happening?
A. The biggest reason is that algorithms learn more easily from behaviors that are immediately visib...
Q. It seems like the key of this study isn’t just leaving more notes, but helping the AI understand the flow of thoughts itself. So when does this approach work best in practice, and when does it provide little benefit?
A. The biggest impact is in situations where information is scattered across multiple places and whe...
Q. Could the key benefit of this approach be that it reduces the burden of having to re-explain things in words? If so, I’m curious in what kinds of writing tasks the effect is likely to be strongest.
A. Yes—the core is to prevent users from having to repeatedly translate the feelings in their heads ...
Q. You mentioned that having AI write a business plan quickly like BizChat can be helpful, but it might also reduce the thinking process users need to go through. So how can we preserve both ‘helpful speed’ and ‘slowness that gives time to think’?
A. The key is to make AI a tool that prompts users to think again, rather than one that writes every...
Q. Why does involving students directly in policy-making lead to better rule compliance?
A. It’s easier to accept the rules when students think through why they’re needed, rather than simpl...
Q. Why can interaction with conversational AI cause some people to further entrench delusion-like thoughts? It doesn’t seem like it’s only a problem for the vulnerable—what’s the core principle?
A. The core is that when people face AI, they can easily come to feel strongly that they are actuall...
Q. Do you think it’s possible to assign a single color or shape to each emotion, or do you think there are commonalities even when accounting for differences between individuals?
A. It’s difficult for exactly the same rules to apply to everyone, because people grow up in differe...
Q. Why did students differ even when given the same assignment—some handed everything to AI, while others used AI while correcting it themselves?
A. The biggest reason is that students differ in how they perceive AI and their habits for using tim...
Q. Why can a small praise signal do more than just make people feel good—and actually increase helping behaviors later?
A. Small praise gives people a sense that ‘what I did mattered.’ As a result, they may start to view...
Q. Why might AI broaden the scope of a question, yet still reduce the diversity of the answers?
A. Because AI can generate multiple possibilities at once, the question itself can become broader. H...
Q. Even if GenAI helps in group learning, it might end up weakening students’ ability to think and coordinate on their own. In what situations do you think it helps—and in what situations does it get in the way?
A. The key is that GenAI helps when it plays a role that prompts students to collaborate better with...
Q. The key of this system seems to be not just displaying instructions well, but helping users understand exactly what they need to do and only showing guidance at the right moments. Why is this approach so important?
A. The biggest reason is that people lose a lot of focus when they have to read explanations and con...
Q. You said LLM-NPCs increase players’ burden while also increasing freedom—why do these seemingly contradictory reactions show up at the same time?
A. Because LLM-NPCs produce responses that are far more natural and less predictable than NPCs with ...
Q. Why does an AI agent’s name have such a big impact on making users trust it more and use it longer—more than just sounding nice?
A. Because a name acts like the user’s ‘first impression.’ When people have a name for something, th...