Viewpoint Vault

Social circles, academic tracks, algorithmic media, and classroom norms can quietly produce ideologically homogeneous environments. We focused on upperclassmen because they have had the most time to settle into routines, majors, and friend groups, which makes long-term patterns easier to see.

The problem was never that students disagreed with each other. The more interesting question was the gap between stated values and lived behavior.

That framing pushed the study toward social, environmental, and behavioral conditions rather than opinions themselves. We wanted to understand the conditions around speaking, not the content of what was said.

We began with a literature review and contextual interviews with nine Cornell upperclassmen recruited by word of mouth. The interviews explored awareness of echo chambers, how students seek out or avoid differing perspectives, the role of social groups, media consumption, and comfort discussing polarizing subjects.

Thematic coding surfaced consistent patterns: a preference for like-minded peers, avoidance of polarizing topics, heavy reliance on algorithm-driven news, and limited awareness of how narrow those information environments had become.

Interviews capture what people remember and are willing to articulate. Echo chambers are built out of the things people do not think to report: repeated exposure, habitual media, comfortable conversations, and small moments where someone decides whether or not to speak. A diary study let us observe those moments closer to when they happened.

We recruited nine Cornell upperclassmen through email and peer outreach, all active social media users enrolled in at least one social or political course. Each completed a short structured entry once a day for five days, submitted between 7pm and midnight to encourage end-of-day reflection. In total we collected 33 entries.

The most discussed topics across entries were tariffs and climate change, followed by Cornell funding, AI, universal basic income, minimum wage, and Israel and Palestine. What mattered less was the topic and more the pattern around it: media reinforced beliefs far more often than it challenged them, and comfort tracked closely with agreement.

We ran a rapid design ethnography with four Cornell upperclassmen, shadowing them across three classes and one public speaker session. Two researchers observed each session to keep notes accurate and perspectives plural, documenting participation through typed notes and drawings of the space.

Short unstructured interviews afterward gave participants room to reflect on moments of ideological reinforcement or exclusion, and to co-ideate possible responses with us. We deliberately used the interview guide lightly so the conversation could move where the participant wanted it to go.

One unexpected pattern: humor lowered the stakes of participating. That observation later became the Food for Thought prompts in the design, lighter questions that warm a class up before heavier ones.

Students wanted diverse dialogue. Existing classroom structures simply did not make participation feel low-risk, so the easiest and safest option was to stay quiet.

The prototype is the outcome of the research, not its starting point. It is a two-sided system because the research surfaced two different problems: students needed a lower-risk way to contribute, and instructors needed visibility into perspectives that raised hands never reveal.

Our observations showed that instructors often steer discussion reactively, responding to whoever is willing to speak. The same few hands go up, and the conversation follows them. Viewpoint Vault gives instructors a pre-discussion view of the classroom's wider perspective landscape, so a broad range of views can be referenced without singling anyone out.

Neither side works alone. Anonymous student input without an instructor to use it is a suggestion box, and instructor preparation without honest input is guesswork. The value sits in the loop between them.

It came from noticing the gap between what students said they valued, diverse perspectives, and what they actually felt comfortable doing in social and classroom environments. No one asked for an anonymous pre-discussion forum. The need for one showed up in the distance between stated belief and observed behavior.