Curio
A semester-long research project on a small, everyday failure: the moment something interesting appears in your feed, then vanishes before you act on it. Through sixteen interviews, a five-day diary study, and a two-session co-design, our team of four studied how curiosity emerges and collapses on Instagram, then designed Curio, a low-effort app built to keep that curiosity alive.
The problem isn't a lack of interesting content. It's that curiosity collapses before you can act on it.
We call this interpretive closure: an interesting encounter gets absorbed into a familiar frame, and the user scrolls on. One participant compared it to walking into a room and forgetting why you walked in. By the time the feed moves, the question is already gone. Our research focused on that exact micro-moment, when curiosity could emerge but instead gets short-circuited.
What the research already told us about curiosity and feeds.
Before designing anything, we grounded the project in prior work on how curiosity forms and how recommendation systems shape what people see.
Curiosity is motivation, and it is fragile
Curiosity shapes how people learn and experience the world, and can point toward socializing, opportunities, or risk (Kashdan, 2020). It often emerges for information people already care about, but it fails to activate when the knowledge gap feels too small, too predictable, or closes too fast (Kidd, 2015; Loewenstein, 1994). Some models even score curiosity from novelty and diversity, while noting that people weigh those traits differently (Zhao, 2016).
Algorithms shape what we encounter
Adaptive recommendation systems tailor content to keep users engaged, quietly shaping the media environment people move through (Eslami et al., 2015; Bucher, 2017). Within that flow, users experience information encountering, coming across unexpected content while browsing something related (Erdelez, 1999). Over time these systems shape habits, interpretations, and attitudes, and people lean on them for personal, social, and informational needs (DeVito et al., 2018; Cleofas et al., 2022).
The familiarity trap
A preference for a content format can be mistaken for a preference for a topic, so algorithms over-serve familiar subjects and quietly lower the diversity of a feed (Dean, 2022). To make sense of opaque systems, users build folk theories of how the feed works, which further shape what they expect and choose to engage with (DeVito et al., 2018).
Agency, and the lever for design
When adaptability decides most of what users see, people can lose a sense of agency over their own consumption, and content that does not match the algorithm's standards is rarely surfaced (Rader et al., 2015; Klinger & Svensson, 2018). One promising counter-move is curiosity priming: affordances that nudge people to learn more about something instead of dismissing it (Neubrander, 2026).
Two questions about a moment most people never notice.
Under what conditions does curiosity emerge as people encounter content on Instagram during everyday use?
As that curiosity develops, what does it grow into, in feelings, interpretations, and actions?
A qualitative, human-centered study across four rounds.
Because curiosity is fleeting and easy to forget, we chose methods that could catch it close to the moment it happened, in people's own words and their own behavior.
Contextual interviews
8 students on media habits, identity, and mindfulness. Synthesized on a FigJam vision board.
Diary study
5-day autoethnography where each of us deliberately disrupted our own feeds and journaled the result.
Contextual interviews
8 more students, narrowed from general media behavior to the moments curiosity emerges and collapses.
Co-design
Two workshop sessions where participants annotated a mockup and brainstormed features with us.
Sixteen students, two rounds, mapped on FigJam.
We recruited Cornell undergraduates through Instagram story posts linked to a Google Form, collected voice memos, transcribed them, and synthesized the transcripts into a FigJam analysis board to map patterns across responses. Round 1 established how students move through curated feeds; Round 2 narrowed in on curiosity itself.
For five days, we broke our own feeds on purpose.
Each team member acted as both researcher and participant. Because a curated feed rarely surfaces the unfamiliar, we deliberately disrupted it, then journaled the immediate reaction in a daily Google Form. The entries were analyzed with thematic analysis on FigJam.
Ask each other for an unfamiliar topic, search it on Instagram, watch at least 3 posts fully, then scroll normally, journaling every moment curiosity emerged or collapsed.
Immediate emotional responses to unfamiliar content, whether curiosity grew or died, actions like saving or sharing, and whether we kept exploring the topic.
We put participants in the designer's seat.
Across two sessions, eight undergraduates worked in small groups to annotate our prototype mockup, brainstorm curiosity-supporting features on a whiteboard, and reflect on their own scrolling. Giving them a tangible mockup to react to lowered creative friction and surfaced how users themselves would structure exploration. Photos from the sessions are below.
Ideas the participants gave us, and where they landed.
Save into collections, not a single pile
Participants wanted to sort saved videos into named collections, with an auto-suggested album and a Pinterest-style grid to browse them. This fed directly into the Saved and Rabbit Holes features.
Show an example, do not list topics
Instead of listing topics, which they felt broke scroll engagement, participants proposed an example reel for each category. If someone watches the last few seconds or likes a reel, that topic becomes their chosen lane, feeding a "Deep Dive" mosaic of related videos. This shaped Explore and Pathway.
Keep exploration light
Annotating our mockup, participants marked the node graph as "too complex" and "cluttered," and asked for a lighter nudge: a prompt showing content they are likely to watch, or being taken straight to reels. It kept us honest about staying low-effort.
Refine the node map
On the map itself, they suggested node popups that recommend where to add a topic, a larger circle for the main category, randomized topic pairings, and links out to articles or newsletters for going deeper.
Curiosity is fragile, and it needs support at the exact moment it appears.
1. People don't open the feed to learn.
Participants described scrolling as a way to fill downtime, manage boredom, or avoid being alone with their thoughts. When social media is opened for relief, unfamiliar content rarely produces curiosity on its own, and heavy news could even turn early curiosity into anxiety, prompting people to close the app.
"Scrolling is my favorite time of the day, because you don't talk to anyone."
Participant, Round 2 interview
2. Curiosity appears, then collapses before you act.
Across interviews and the diary study, curiosity often flickered and disappeared before participants could do anything with it. Feeds move fast, and a new post replaces the interesting thought before it can be explored. Saving helped a little, but it usually delayed curiosity rather than sustaining it. The takeaway: curiosity needs support when it first appears, not only after it has already been postponed.
"It's like walking into a room and forgetting why you walked into the room."
Jin, diary + interview
3. More context helps, but only if it stays close and low-effort.
Co-designers described real curiosity as a "slow rush," gradual rather than instant, but they did not want exploration to require extra steps. Unfamiliar content was most likely to sustain curiosity when it connected to something familiar and stayed inside the same flow. Total disruption created confusion, not interest.
"Explore. That's like a whole nother step people have to think about what they want to see."
Co-design, Session 1
4. Curiosity is personal.
What sustained curiosity was rarely the topic itself, it was personal relevance. Content kept people engaged when it connected to their identity, interests, prior experiences, or current concerns. For some that meant cultural familiarity or an emotional resonance; for others, a tie to travel, a career goal, or an existing hobby. The same post could spark one person and slide right past another.
5. The spark is visual and emotional before it is intellectual.
What interrupted the scroll in the first place was intensity, not information. Content that was aesthetically striking, surprising, funny, urgent, or emotionally charged was far more likely to stop a thumb and open a moment of curiosity. Interest usually arrived as a feeling first, and only then became a question worth pursuing.
6. Much of the curiosity we cared about was invisible.
When curiosity was sustained, it sometimes surfaced as visible action: opening comments, replaying a post, visiting a creator, or searching elsewhere. Just as often, it developed inwardly. People reflected, imagined, or made a personal association, then moved on without ever touching the screen.
This reframed the whole project: a lot of real curiosity never shows up as a tap, so designing and measuring for clicks alone would miss most of it. It is a big reason Curio tracks the journey through Pathway and Rabbit Holes, not engagement counts.
Curio, a low-effort app that keeps curiosity moving.
Familiarity is built into the structure and the interface, while the unfamiliar shows up in the content and the connections users can explore, so going deeper never feels like a separate task. The home feed is short-form video with familiar actions (explore, like, comment, bookmark, share) and a feed-settings control for cognitive load.
Explore
A "curiosity feed" shown as a node-graph of topics and their connections. Zoom in to pick subtopics; each node blends roughly 60% familiar and 40% unfamiliar content, so novelty arrives through a structure that already feels recognizable.
Pathway
A visualization of the route between topics: the videos, subjects, and time spent along the way. Paths can be shared with friends, and an "explore more" option extends the trail so novel content arrives from a familiar direction.
Rabbit Holes / Saved
Stats on what you recently engaged with and how an interest grew over time, with nudges that bring you back to saved videos through notifications instead of letting them be forgotten.
Feed settings
A lightweight control to adjust cognitive load, so the feed can meet people where they are, whether they want relief or a slow rush of something new.
Move through the curiosity flow.
An interactive Figma prototype of Curio's core loop, from a moment of interest into deeper exploration without breaking the flow of scrolling.
The conclusion is qualitative by design.
What we learned
Curiosity collapses when unfamiliar content asks for too much effort, and it continues when unfamiliarity is balanced with a familiar presentation. Rather than chase an exact ratio, we built familiarity into the structure of each item and the interface, and let the unfamiliar live in the content and the connections people can explore.
We hoped to ship a fully adaptive recommendation engine but ran into time and technical limits, so we prioritized the interaction flow instead, since reducing resistance is an important part of sustaining curiosity. That is the first thing I would build next.
Want to talk through the research?
Happy to walk through the methods, the synthesis, or how the findings shaped every feature in Curio.