Before this class, I thought that product design and user research was a relatively simple process: talk to people, figure out what they want, and then build something in line with what they want. It was enough to just understand that people had a problem, then come up with a solution to that problem by asking them, and then build that out. The main thing I didn’t realize is how the complexity of us as humans and our psychology makes that such a complicated and multifaceted task, and how much research, investigation, and testing is required when it comes to designing products for behavior change.
During the class I really enjoyed Christina’s lectures on user psychology, crafting intervention solutions, product design concepts like friction, and just the slides in general that helped me develop product intuition for things that work and solve people’s problems. In particular, I really enjoyed the moments in class that gave me a deeper appreciation for the level of research and thought that goes into truly groundbreaking products. There are a lot of times that software interfaces can feel “good” or “enough”, but there are only a few times where there’s a truly exceptional experience. In this class, I think I realized that what makes an exceptional experience vs. a “good enough” one is just a level of user understanding and product intuition built on a much stronger and deeper foundation of domain knowledge and user research.
The parts I didn’t like that much were a lot of the diagramming and outlining frameworks since they felt a lot less intuitive and useful. Some examples were the affinity maps/system paths/bubble maps. In some situations it felt like we were diagramming just to diagram and we weren’t actually creating new ideas and thinking that deeply. I think part of the reason for this is sometimes Christina would just introduce a new map/diagram format and we’d be expected to have something done in like 5-10 minutes. In a lot of these situations my team just didn’t have the mental energy/see the need to dig beyond the surface and use the map as a framework for product improvement, so we would often just repackage stuff we already know and figured out a long time ago into a different format. That’s why I feel like it would be so much better if the class had deep dives on certain product mapping/drawing frameworks, with more detailed examples of their functionality and use cases (and maybe case studies of real successful products or product features). That would make it so much more interesting to actually think about how to apply those techniques to our own products, rather than just making us adapt to a different drawing/mapping technique in such a short timeframe.
I honestly wish that I spent more time in the ideation phase trying to figure out what to build, because I feel like even though we targeted an interesting problem the solution space was pretty limited and sort of pigeonholed into one format of intervention, which would be fine but it felt more restrictive and I don’t think we could be as creative with some parts of the intervention.
In terms of our project JACE, the main nudge we use is we intercept a user’s first prompt to an LLM, ask them reflection questions about it, and basically make users use more critical thinking and reflect a little more on the topic prior to allowing the prompt to be sent to the LLM, both enhancing the user’s own thinking processes and also updating the prompt with more context (thus making the resulting output higher quality). We believed this was a pretty universally good nudge that wouldn’t have many adverse/unforeseen consequences to users. Because we intervene before users interact with an LLM, there’s much lower risk of accidental manipulation because we don’t influence or alter the LLM output in any way (besides appending additional context from the user’s own thinking to the user’s initial prompt). Essentially, this means that our implementation functions one layer of abstraction higher than the actual conversation, so either users end up using their critical thinking and reflect more or they just uninstall if it becomes too much of a bother. The main reason we support privacy is because collecting more of the user data does not improve the quality of the product and the product will not rely on the user data (of their first prompts to LLMs) to function or survive. In the future, it is possible that there will be a more advanced implementation of our solution (which is to prompt/intervene at certain moments in the chat when appropriate, not just at the initial message) which could pose more privacy concerns since we would need to be continually reading the conversation messages.
Now I think that design for behavior change is a much more complex and multifaceted process than I once thought. Every time you build a product you have to consider all the stakeholders/potential users, ethical concerns and privacy risks, and you have to actually do extensive user research to figure out what people want in the first place. One of the biggest realizations and changes in thinking I had was realizing how much more complicated understanding what intervention to build is. It’s a very long and difficult process to understand users at scale, and it takes a lot of time to know how to ask the right questions and figure out the most important details.
In the future when I’m faced with figuring out how to design an effective intervention to an important problem, I will approach it from a more deeper foundational lens of understanding user psychology and crafting a product that balances the fine line between different users, accessibility needs, ethical and privacy risks, and also just behavior design effectiveness. I will better understand the importance of using drawing and feature diagramming frameworks to look at the foundational motivations of behavior. I think the distinction between behavior as the result of individual actions and thought processes vs as a result of environmental and external conditions is really important to note as well, in the sense that interventions have to be effective implementations and integrations of routine in daily life.
