A Reflective Guide to Generating Good Ideas in HCI and HCAI Research
Good Ideas and Where to Find Them?
A good idea isn’t just novel—it’s something you care enough about to sit with for a while. This reflection offers a practical, personal framework for generating HCI research ideas that matter. Whether you’re stuck or overflowing with possibilities, this guide helps you generate ideas and trace where the novelty lies.
What is a good idea?
A good idea is something you can dedicate time to without losing interest. It’s deeply personal. What feels exciting and meaningful to one person might feel like torture to someone else. For example, the idea of building and validating questionnaires doesn’t excite me at all. Still, some researchers devote years to doing just that (and some genuinely enjoy it — believe it or not).
For me, understanding how humans form mental models of conversational AI is endlessly fascinating. Others might disagree. That’s okay. A good idea is one that you can get excited about — and more importantly, one you can get others excited about too. That might include collaborators, advisors, reviewers, and even your future self.
Where do good ideas come from?
That’s the tricky part. The good news? Sometimes, good ideas just come to you.
During a user study with Amazon Alexa, one participant encountered an error, laughed, and said, “It’s like a silly child.” That metaphor stuck with me. I ended up dedicating the next three years to studying why people use metaphors to describe voice interfaces, what those metaphors reveal, and how we might use them to inform design [4, 5, 7].
Sometimes, people will hand you great ideas without even realizing it. I’ve found this happening more often as I’ve progressed in my career. I have more ideas than I have time for. So when students or colleagues reach out, I’m often happy to share an idea that excites me — if it excites them too, and they have the bandwidth to run with it.
Ideas can come from anywhere: in conversation, during a walk, in the shower, while teaching, while reading literature, while watching someone interact with technology in a way you didn’t expect. The key is to put yourself in situations where ideas can find you. Go to talks. Read papers or blogs every day. Reach out to researchers whose work you admire. Schedule coffee chats or Zoom calls. Make yourself available to curiosity.
But what about when the ideas stop coming?
That’s harder. In 2020, right after I completed my PhD candidacy exam, I was ready to launch my own projects. But COVID hit. No one was meeting. We were all isolated and doom-scrolling. It was difficult — nearly impossible — to be in the kinds of situations where ideas typically spark.
COVID is a rather extreme example, but often even when you do all the right things, inspiration still doesn’t come. That’s normal. But for those moments, I’ve found it helpful to fall back on a structure that can guide toward idea generation.
Here’s the framework I’ve come to rely on. You can think of it as a fill-in-the-blanks formula:
An innovation in [something], designed for a [user community], situated in a [context], to understand or support a [goal].
Before we begin applying the framework, it’s worth taking a moment to unpack what each component actually means. Terms like “something” or “context” might appear deceptively simple, but in HCI they often carry layered and situated meanings. By clarifying each element — what we build, who we design for, where the interaction takes place, and what we’re ultimately trying to achieve — we can better understand how these dimensions shape the kinds of ideas we pursue. Once we’ve defined these building blocks, I’ll walk through examples from my own work and others to show how the framework comes to life in practice.
Defining “Something”
Let us begin by examining the first component in the idea generation framework: “something.” At first glance, this may seem self-evident — perhaps a piece of technology, a new method, a clever interface. But within HCI, this notion quickly becomes less stable. What counts as a “thing” in HCI is not just a matter of artifact — it is entangled with theory, method, and practice. And unlike engineering disciplines that privilege reproducibility or optimization, meaning-making in HCI often derives from interpretive work, contextual relevance, and social implications [13].
Indeed, even within our own community, the “thingness” of an HCI contribution varies: some papers propose new tools or systems, others new methods or techniques, and still others offer a conceptual shift — a reframing of how we understand a phenomenon. None of these are necessarily more valid than the other; rather, they emerge from different epistemic traditions. As Blackwell reminds us in his writing on metaphor, the act of turning ideas into design tools — what he calls reification — is itself a sociotechnical process [3]. This applies equally to how we turn systems, methods, or theories into the “something” that forms the seed of an idea.
Let’s consider artifacts first. In HCI, artifacts might include a novel interface, a voice assistant, a speculative prototype, a chatbot prompt structure, or a reimagined design probe. It does not always have to be something you’ve built from scratch — studying an existing system in a new way, reappropriating a commercial tool for a new purpose, or rethinking how users engage with emergent technologies (such as LLMs) can all count as innovation. The key is novelty: your artifact must push boundaries, even if incrementally.
Similarly, theoretical contributions can take many forms: you might propose a new conceptual framework, redefine an existing construct, or adapt a theory from another discipline to an HCI context. These, too, are “things” — but abstract ones. They are often the quiet backbone of a paper, giving structure to how we think and language to what we notice.
Then there is method. Methodological contributions are particularly interesting because they shape not just what we learn, but how we come to know it. A new method might involve a novel combination of techniques (involving qualitative and/or quantitative data), or an adaptation of traditional methods to suit the research question. Even small additions — such as the use of persona-creation as a co-design activity for relational conversational agents [10]— can constitute meaningful innovation.
Of course, these categories are rarely cleanly separated, but they don’t need to be equally emphasized either. Some papers introduce innovation in just one thing, which can be enough. Others manage to span multiple dimensions, layering methodological, theoretical, and artifact-level novelty into a tightly integrated contribution. These are often the papers that leave a lasting mark.
For instance, in one of my own papers [6], we developed a voice assistant that delivered a physical activity program to older adults. Here, the primary contribution was in the artifact itself — a conversational agent that combined NIA-approved elder-friendly exercises through a carefully designed voice persona. It was designed with care and grounded in real needs, but the innovation was largely at the level of the system.
Contrast this with Porcheron et al.’s work on Voice Interfaces in Everyday Life [14], one of my favorite papers in the field. Their study of smart speakers (then an emerging technology) combined underutilized methods (ethnomethodology and conversation analysis), and a subtle theoretical rethinking of how everyday voice interaction unfolds. It was an ambitious contribution across all three things — artifact, method, and theory — and it shows.
So while you can build a strong paper around innovation in just one area, the most influential work often brings these threads together in ways that challenge how we build, how we study, and how we think.
Defining “User Communities” and “Contexts”
Compared to methods or theories, defining the user community and context tends to be more straightforward — at least on the surface. HCI as a field places users at its core. But users are not generic. While each individual brings their own lived experience, it is often analytically useful to group users into communities that share certain behaviors, needs, or experiences.
These communities can be shaped by demographics — such as age (older adults, children), gender, race, or disability status. But they can also be defined by practice, affinity, or expertise: clinicians in emergency departments, generative AI users, VR enthusiasts, educators, YouTube creators. Some communities form around shared goals or tools. Others are united by marginalization or systemic exclusion. And they don’t even have to be human — animal interaction, robotic companions, and AI as a social actor all raise questions about who counts as a “user” in design.
Crucially, users don’t exist in a vacuum. Every interaction is shaped by the context in which it occurs. Context might refer to a physical setting (a hospital waiting room, a public park, a home office), but it also includes the activity the person is engaged in (conversing, browsing, grieving, learning). Context introduces constraints, norms, noise — and often, insight.
From a research perspective, novelty can emerge in several ways:
-You might study a well-known user group in a new context — for example, examining how people interact with voice assistants in public social settings like pubs or cafés. [15]
-You might study an underexamined user group in an everyday context , such as how parrots use video calling systems in a home environment. [12]
-Or you might focus on an understudied user group engaged in a distinctive context —how low-income South Asians use public smart speakers in urban slums. [16]
The richness of HCI often lies in pairing users with places and practices in unexpected ways. The framework I propose invites you to consider both who is being studied and where and how they experience the systems we are interested in.
Defining the Research Goal
In HCI, research often orients itself around two broad trajectories: understanding people’s interaction with technology, or supporting those interactions in some meaningful way. The distinction is useful, but not always clean. Understanding and support are often entangled, and many good papers shift between them fluidly.
Understanding, in this sense, is not just about describing behavior. It’s about making sense of the mental models, cultural logics, constraints, and tacit expectations that people bring into their interactions with technology. These moments are often most visible when things don’t go as planned — when the user’s actions exceed the system’s boundaries, or when the system responds in ways the user did not anticipate. A visually impaired person navigating a touchscreen kiosk with no audio feedback. A clinician ignoring a pop-up alert they’ve learned to mistrust. A teenager anthropomorphizing a chatbot out of loneliness. These are not bugs in the system; they are moments where systems and people talk past one another.
In most cases, these moments are unintentional. Other times, they are designed that way. Systems that nudge users toward endless scrolling or make the “unsubscribe” button intentionally hard to find also deserve study — not just to critique them, but to understand what kind of logic they encode and its effect on people. Understanding, then, is rarely neutral. It can be an act of documentation, but also of interrogation.
Support, on the other hand, takes on a more generative stance. It asks what kinds of interventions — technological, methodological, or theoretical— might make people’s lives easier, richer, or more meaningful. This could mean designing new tools, refining interactions, or proposing new ways of seeing a problem that eventually guide better design. But it doesn’t have to be about “solving” anything. Support can be small. It can be about attending to a previously overlooked experience, or amplifying a quiet but meaningful practice. And sometimes, support begins by first committing to understand.
What’s considered worth supporting, or even worth understanding, is deeply personal. Some researchers care about accessibility, some about play, some about infrastructure. Some write code, others write critiques. The work becomes interesting when the goal is not just well-intentioned, but clearly motivated — when it is animated by a sense of care, curiosity, or even frustration. The research goal is not just the endpoint; it’s the orientation from which all the other decisions follow.
Applying the Framework
When ideas aren’t coming naturally, this framework gives me something to hold onto. But like any good structure, it’s not a formula — it’s a scaffold. And it works best when you begin not with the artifact, or even the user, but with the goal. Not just “build an app” or “conduct a study,” but the deeper question: what do I care enough to sit with for a while?
Start with What Matters
Let’s say your goal is helping people better understand complex health information. That’s still broad, so the next question becomes: who does this matter to most? You might think of older adults navigating a new diagnosis, or family caregivers trying to make sense of conflicting treatment plans. The user community begins to take shape.
But people don’t interact with information in a vacuum — they do so somewhere specific. At home with a smart speaker murmuring from the kitchen. In a waiting room with paperwork and a flickering television. Or on a quiet afternoon at the kitchen table with tea. Context narrows the field and shows you where this problem actually lives.
Once you have goal, user, and context, the final question becomes: what might make a difference here? This is where your “something” emerges. It might be a new kind of conversational agent, a different way of framing information, or even a method for understanding what users misunderstand and why. It doesn’t need to be revolutionary — but it should shift something: what’s possible, visible, or thinkable.
Let’s go back to the framework prompt to see how the idea materializes. A helpful trick is to imagine you’re filling in the sentence“An innovation in [something], designed for a [user community], situated in a [context], to understand or support a [goal].”
Seeing the Framework in Action
In one of my studies, this process led to designing a voice assistant that shared health information through a serialized murder mystery. The goal was engaging health learning. The users were older adults. The context was the home, where people interact with technology informally and often alone. The innovation? Combining narrative scaffolding (something: theory) with proven self-regulated learning strategies (something : theory) using a smart speaker-based voice assistant (something: artifact)— an unexpected but resonant pairing.
Translated into our framework as: An innovation in combining narrative scaffolding and self-regulated learning strategies through a smart speaker-based voice assistant (something), designed for older adults (user community), situated in the home where technology is used informally and often alone (context), to support engaging health learning (goal).
When you read HCI papers with this framework in mind, you’ll start seeing this logic everywhere. Here are examples from some influential papers:
An innovation in a conceptual framework of feminist design qualities (something) designed for HCI researchers and practitioners (user community) situated in the critical turn in HCI (context) to support more inclusive, reflexive, and socially just design practices (goal). [2]
An innovation in the use of collaborative online games for distributed human computation (something) designed for everyday Internet users seeking entertainment (user community) situated in short online gameplay sessions with randomly paired partners (context) to support large-scale labeling of web images for search, accessibility, and machine learning applications (goal). [1]
An innovation in adding social context to AI explanations (something) designed for people using AI in high-stakes decisions (user community) situated within everyday team workflows (context) to support users in making informed, accountable choices (goal). [8]
An innovation in analyzing user frustration and recovery strategies in chat‐based LLMs (something) designed for end users collaborating with AI on tasks like writing, coding, or research (user community) situated during iterative conversational sessions where users refine AI outputs (context) to understand why users feel dissatisfied and guide improvements in LLM usability (goal). [11]
An innovation in mixed-initiative user interfaces (something),
designed for people collaborating with intelligent assistants on tasks like scheduling and information lookup (user community) situated in interactive, email-based scheduling tools where control can shift between user and system (context) to understand how dynamic control shifts influence conversation efficiency and user agency (goal). [9]
Here, the novelty will always lie in something and might also lie in the pairing of user and context, or it may simply be a goal that’s never been studied quite this way. The point isn’t to fill every blank with a grand revelation — it’s to trace where the energy lies. The strongest papers usually hinge on at least one element being unmistakably fresh. The very best ones connect all four in ways that make it impossible to imagine one without the others.
When You’re Stuck
So if you’re stuck, start with the goal. Let it lead you to the people who care. See where they live, work, pause, and struggle. Then ask yourself: what small twist — what something — might you offer that no one has quite tried before?
The framework works because it transforms the overwhelming question “What should I research?” into a more manageable conversation between four interconnected pieces. Not every piece needs to be revolutionary, but together they should point toward something worth your time — and worth others’ attention too.
References
1.Luis von Ahn and Laura Dabbish. 2004. Labeling images with a computer game. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’04), 319–326. https://doi.org/10.1145/985692.985733
2. Shaowen Bardzell. 2010. Feminist HCI: taking stock and outlining an agenda for design. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10), 1301–1310. https://doi.org/10.1145/1753326.1753521
3. Alan F. Blackwell. 2006. The reification of metaphor as a design tool. ACM Trans. Comput.-Hum. Interact. 13, 4: 490–530. https://doi.org/10.1145/1188816.1188820
4. Jessie Chin, Smit Desai, Sheny (Cheng-Hsuan) Lin, and Shannon Mejia. 2024. Like My Aunt Dorothy: Effects of Conversational Styles on Perceptions, Acceptance and Metaphorical Descriptions of Voice Assistants during Later Adulthood. Proc. ACM Hum.-Comput. Interact. 8, CSCW1: 88:1–88:21. https://doi.org/10.1145/3637365
5. Smit Desai, Jessie Chin, Dakuo Wang, Benjamin Cowan, and Michael Twidale. 2025. Toward Metaphor-Fluid Conversation Design for Voice User Interfaces. https://doi.org/10.48550/arXiv.2502.11554
6. Smit Desai, Xinhui Hu, Morgan Lundy, and Jessie Chin. 2023. Using Experience-Based Participatory Approach to Design Interactive Voice User Interfaces for Delivering Physical Activity Programs with Older Adults. In Proceedings of the 11th International Conference on Human-Agent Interaction (HAI ’23), 180–190. https://doi.org/10.1145/3623809.3623827
7. Smit Desai and Michael Twidale. 2023. Metaphors in Voice User Interfaces: A Slippery Fish. ACM Transactions on Computer-Human Interaction 30, 6: 89:1–89:37. https://doi.org/10.1145/3609326
8. Upol Ehsan, Q. Vera Liao, Michael Muller, Mark O. Riedl, and Justin D. Weisz. 2021. Expanding Explainability: Towards Social Transparency in AI systems. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (CHI ’21), 1–19. https://doi.org/10.1145/3411764.3445188
9. Eric Horvitz. 1999. Principles of mixed-initiative user interfaces. In Proceedings of the SIGCHI conference on Human Factors in Computing Systems (CHI ’99), 159–166. https://doi.org/10.1145/302979.303030
10. Xinhui Hu, Smit Desai, Morgan Lundy, and Jessie Chin. 2024. Beyond Functionality: Co-Designing Voice User Interfaces for Older Adults’ Well-being. https://doi.org/10.48550/arXiv.2409.08449
11. Yoonsu Kim, Jueon Lee, Seoyoung Kim, Jaehyuk Park, and Juho Kim. 2024. Understanding Users’ Dissatisfaction with ChatGPT Responses: Types, Resolving Tactics, and the Effect of Knowledge Level. In Proceedings of the 29th International Conference on Intelligent User Interfaces (IUI ’24), 385–404. https://doi.org/10.1145/3640543.3645148
12. Rebecca Kleinberger, Jennifer Cunha, Megha M Vemuri, and Ilyena Hirskyj-Douglas. 2023. Birds of a Feather Video-Flock Together: Design and Evaluation of an Agency-Based Parrot-to-Parrot Video-Calling System for Interspecies Ethical Enrichment. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (CHI ’23), 1–16. https://doi.org/10.1145/3544548.3581166
13. Elisa D. Mekler and Kasper Hornbæk. 2019. A Framework for the Experience of Meaning in Human-Computer Interaction. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems (CHI ’19), 1–15. https://doi.org/10.1145/3290605.3300455
14. Martin Porcheron, Joel E. Fischer, Stuart Reeves, and Sarah Sharples. 2018. Voice Interfaces in Everyday Life. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ’18), 1–12. https://doi.org/10.1145/3173574.3174214
15. Martin Porcheron, Joel E. Fischer, and Sarah Sharples. 2017. “Do Animals Have Accents?”: Talking with Agents in Multi-Party Conversation. In Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing (CSCW ’17), 207–219. https://doi.org/10.1145/2998181.2998298
16. Simon Robinson, Jennifer Pearson, Shashank Ahire, Rini Ahirwar, Bhakti Bhikne, Nimish Maravi, and Matt Jones. 2018. Revisiting “Hole in the Wall” Computing: Private Smart Speakers and Public Slum Settings. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ’18), 1–11. https://doi.org/10.1145/3173574.3174072