let attendees continue the conversation

Debates, panels and moderated dialogues invite experts to model constructive disagreement for an audience. Sway gives that audience somewhere to take it. It is one thing to observe constructive disagreement; it is another to practice it yourself.

~1 min for an attendee to register opinions
5 min of programme time, in the room
56% at a polar extreme moderated
86% found the facilitator even-handed

how an event deployment works

Before the event, you choose the discussion topics. Three statements are enough to produce good matches in a large audience.
At the close of the event, attendees scan a QR code and register their opinions. This takes about a minute.
Over the following days, Sway matches attendees who disagree and invites each pair into a one-on-one conversation.
Conversations stay open in the weeks that follow. Attendees chat from their phones or laptops, on their own schedules. You choose the window — a few days, the rest of a themed week, or longer.

Set aside about five minutes in the programme for opinion registration, while the audience is still in the room. Project the QR code, explain the invitation, and give attendees a minute to respond on their phones. Experience across events is consistent: once the audience disperses, participation drops sharply — however good the flyer or the follow-up email.

Once opinions are registered, Sway handles the logistics. It notifies each attendee when they are matched, helps the pair agree on a time to chat (or to exchange messages asynchronously), and sends reminders as the conversation progresses.

Setup is light: you provide the topics, and the Sway team configures the event and provides a custom link and QR code to project on screen or print on table cards. Sway is grant-funded. Attendees can volunteer to take part in IRB-reviewed research on civil discourse.

choosing topic statements

Good event topics are short, declarative statements that split the room and pick up the themes attendees have just watched discussed on stage. For a debate or dialogue on polarization, for example:

worked well on stage nights

Congress should demographically represent the United States: if the country is 51% women, Congress should be 51% women.
Disliking the opposing political party is now a bigger problem than our disagreements over policy.
Most Americans have a reasonably accurate picture of the political views of the opposing party.

Statements about campus speech and protest are evergreen: students hold real views on them and can engage without technical background. The Sway team is happy to workshop statements with you; the topic guide collects the ones that have worked well, with the split each produced.

what the event report shows

After the discussions conclude, you receive a de-identified event report — behavioural evidence of your event's impact, rather than attendance and satisfaction numbers.

what you get	what it tells you
The arguments, not just the positions	What your audience believed and why: the reasoning on each side, where they found common ground, and which disagreements survived the conversation.
Opinion distributions per topic	With before-and-after shifts for attendees who completed surveys. Pre/post measurement is already built into the platform; adding it to your event requires no extra setup.
Where the reasoning held and where it slipped	Including recurring misconceptions worth addressing from the stage next time.
Follow-ups and next topics	Ready-to-use suggestions for your next event, drawn from what this audience actually argued about.

Browse real reports →

where Sway has been used

Sway has run at a range of public events, including several MIT Free Speech Alliance debates, events at UNC Chapel Hill's School of Civic Life and Leadership, a Committee for Academic Freedom panel in the UK, and a variety of Australian Dialogues. Among them: MFSA's Great Climate Debate and its fall 2025 debate at MIT, and public dialogues in Brisbane and Melbourne on tariffs and trade, the energy transition, the China–Australia relationship, and Taiwan.

does it change minds?

nearly 5×: Participants are nearly five times more likely to moderate an extreme view than to harden a moderate one. 56% who entered a conversation at a polar extreme left holding a more moderate view.
51.5%: The effect holds on the hardest topics — abortion, immigration, gender — where 51.5% still moderated.
−23%: The share holding extreme positions dropped by 23% after a single 30-minute conversation with a partner who disagreed.

From 10,000+ matched before/after opinion pairs at 100+ institutions. Learn more about improving speech cultures →

inclusion & fairness

Guide does not offer opinions. It pushes participants to reason more carefully, interrogate their assumptions, and engage seriously with the strongest version of the opposing view.

In a survey of over 5,000 participants, 86% agreed that Guide treated both sides with equal respect — and only 3% disagreed. In a separate study, more than 600 politically diverse Americans read real Sway transcripts across multiple topics and perceived no directional bias, while correctly detecting bias when it was experimentally introduced.

Students know that Guide is AI and they find it not at all partisan, not at all biased. That was never an issue.

Diana Mutz
Samuel A. Stouffer Professor of Political Science & Communication, University of Pennsylvania; member of the National Academy of Sciences, and author of Hearing the Other Side

Learn more about bias evaluation → · more from instructors

privacy

Conversations are encrypted. Organizers cannot read identifiable transcripts or view any individual attendee's opinions. Personal attacks and mere insults are blocked. Attendee data is never used to train commercial AI models. 91% of participants felt comfortable sharing honest opinions, and 92% felt respected by their partner. Read the privacy and consent document →

Up next

Running Sway at your event, and the material to take to whoever approves it.