Once people see the votes, the vote stops telling you what they think.
When a live poll shows results early, later answers drift toward whatever is winning. You see the final tally and take it as the group's judgment, but part of that tally is people following each other. Answerama is building a way to measure that drift as it happens and to show you what the room thought on its own.
The drift is measurable
In a classroom clicker study, students who saw the class bar graph were 30% more likely to switch to the most common answer. (Perez et al. 2010)
On a news site, one fake early upvote raised final ratings by 25% on average. (Muchnik, Aral & Taylor 2013)
Seeing others' answers can also help. In one experiment, groups that shared estimates in decentralized networks got more accurate. (Becker, Brackbill & Centola 2017)
So the goal is not to block influence. It is to tell when the tally reflects herding and when it reflects people learning from each other. Today a host has no way to tell which happened.
Vote first, then see the room
Step 1. A hidden first vote.
Everyone answers before any results appear. That gives us each person's independent answer for every question.
Step 2. Results, then a revote.
The room sees the tally and can vote again. Comparing the two votes shows how much answers moved and in which direction.
Step 3. Spot the herding.
We train models to estimate, during a normal session, how much the results on screen shifted the answers. The models use timing, answer order, the tally each person saw, and changed votes.
Step 4. Report a corrected result.
Next to the raw tally, Answerama shows an adjusted "independent judgment" result with a margin of uncertainty, sized for groups of about 10 to 300. The method builds on crowd research such as the "surprisingly popular" answer (Prelec, Seung & McCoy 2017).
We don't yet know whether herding can be detected reliably in groups this small and sessions this short. That is the question the research is meant to answer.
What the room saw
What the room thought first
Run a pilot session with your group
We need 60 to 100 live sessions with adult groups: college classes, professional training and workshops. A session is one ordinary meeting with your group, using 8 to 10 questions. Players scan a QR code and join on their phones, with no app or account. Each question runs a hidden first vote, then shows results and allows a revote.
You get your session results and free Answerama Pro for as long as the study runs. Pilot sessions start once ethics review with our academic partner is in place.
Who's building this
Answerama is made by Novemus, LLC, a small business in Harlingen, Texas. Founder Cesar Riojas is a UX designer and U.S. Air Force veteran. He designed and built Answerama's live rooms, TV casting, survey authoring and hosting, and he leads this research.
How we handle pilot data
Players join with a random session ID. There's no app to install and no account to create, and we don't ask players for names or emails. For each question we record the answer, when it came in, and whether the player changed it.
No pilot session runs until ethics review with our academic partner is in place. Hosts see their own session results. Published findings use only de-identified totals, never individual answers. Raw session data is deleted after 24 months, and only de-identified totals are kept after that.
Host sign-up details are used only to run the pilot.