experimental data

We are using Sway to study whether AI facilitation can help people talk productively across deep political and moral divides. In an early pilot study, 120 paid participants each spent 30 minutes on Sway discussing a polarizing topic with a partner who disagreed with them. Before and after the discussion, they rated their agreement with nine statements measuring openness to opposing viewpoints. We are planning a large randomized controlled trial that will compare discussions like these with unmediated ones, and a field study reaching students at over 100 colleges and universities in 2026.

120paid participants

30 minone discussion, one polarizing topic

2 in 3scored higher on openness afterward

+0.32 SDaverage gain on the openness scale

Openness scale: nine statements rated from −3 (strongly disagree) to +3 (strongly agree) before and after the discussion (Cronbach's α = 0.89). On every statement, the average shift after discussion was statistically significant and in the direction of more openness.

The conversations

30 min

of AI-facilitated discussion with a partner who disagreed. Guide, the AI facilitator, opened every chat by introducing the topic and announcing each participant's position. Topics: the 2020 national U.S. election, the January 6 Capitol riot, Israel-Palestine, U.S. funding for Ukraine, and U.S. abortion policy.

Screenshot of a Sway chat in which the AI facilitator, Guide, introduces a discussion on the topic of abortion policy and announces the two participants' positions. — Guide opens a pilot-study discussion of U.S. abortion policy.

In their own words

Participants emerged from discussions of extremely polarizing topics with feedback that consistently amazed us. The comments here are verbatim from post-chat feedback, each labeled with the topic the pair discussed.

This is probably the only way I would ever have a conversation like this. I really appreciated that I could ask Guide for fact-checks instead of saying ‘You are wrong’. The fact that it was mediated helped the tone to stay civil and moved us along.

the 2020 national U.S. election

I definitely think Guide was very useful in leading the discussion. When we got off track, Guide made sure to point us in the right direction. I think this study will be very useful

the January 6 Capitol Riot

In a way, I felt safe with guide. I'm sensitive, so having that slight security with the guide, I felt better to speak my mind a little more.

Israel-Palestine

I thoroughly enjoyed it. The guide was a great mediator that really help set the pacing and also control the discussion. Without the guide, I can see the conversation get messy with people talking over one another. It created a thought provoking conversation instead of two people rambling.

U.S. abortion policy

I feel like this is the best AI-driven anything that I've ever personally experienced. It asked very thoughtful, relevant, specific questions to direct the conversation. It was patient, encouraging, and respectful.

U.S. funding for Ukraine

It was actually amazing. Guide interjected at the right times without being overly intrusive. I went in thinking that I couldn’t compromise, but I left with a more open-minded view. I really enjoyed it.

U.S. abortion policy

I found Guide so useful in developing the discussion further. I think it made both parties delve deeper into some of the statements both made. I really enjoyed discussing this topic with someone else. I love the whole concept.

Israel-Palestine

Not only did Sway guide our conversation in a thought provoking way, it moved us in a very civil direction. I was able to see things from my partner’s point of view. I also changed my stance!

the January 6 Capitol Riot

Understanding the other side

47%

of participants gave a higher rating to “I feel like I can understand people who disagree with me about this topic” after their chat. 14% rated it lower. Agreement rose by 0.41 standard deviations (95% CI: 0.24 to 0.59). The change was highly statistically significant: t(120) = 4.8, p < 0.00001.

Chart comparing participants' before and after ratings of the statement 'I feel like I can understand people who disagree with me about this topic': 47% gave higher ratings after a Sway discussion and only 14% gave lower ratings. — Each blue dot is one participant's rating before or after their chat; grey lines connect the same participant. Red triangles show individual change; the dark triangle marks the average change with its 95% confidence interval.

Crediting the other side

61%

agreed more strongly after their chat that “People who disagree with me about this topic have well-thought out reasons for their views” — more than twice as many as moved in the opposite direction (28%). The improvement was highly statistically significant: t(120) = 4.5, p = 0.0003.

Chart comparing participants' before and after ratings of the statement 'People who disagree with me about this topic have well-thought out reasons for their views': 61% agreed more strongly afterward versus 28% who moved in the opposite direction. — Ratings of “People who disagree with me about this topic have well-thought out reasons for their views,” before and after a Sway chat.

Openness overall

+0.32 SD

average gain on the full nine-item openness scale — a meaningfully large effect. Two-thirds of participants scored higher after chatting with a partner on Sway; one in five scored lower. 95% CI: 0.21 to 0.44; t(120) = 6.1, p < 0.00001. We expect these effects to strengthen and persist after repeated discussions on Sway — a hypothesis we will test in our RCT.

Chart showing the distribution of change on the nine-item openness scale before versus after a Sway discussion: two-thirds of participants scored higher afterward while one-in-five scored lower. — Total scores on the nine-item openness scale (α = 0.89), before and after a Sway chat.

Up next

How these findings hold up in the field, and the material behind them.