I really think there is a strong potential in these things. Don't get fooled by the simplification of seeing opinions in a 2D graph. It helps to explain, but the reality of what these things can (potentially, not sure about this particular implementation) do is to really find across the thousands of dimensions of the debate space, statements that may help you bridge groups.

Imagine person A, strong humanist, no-border, intransigent on human rights. Imagine person B, authoritarian, xenophobic and traditionalist. They are unlikely to agree on statements like "ethnicity X are subhumans" (strong reject by A) and even a middle ground in the form "citizen of ethnicity X should have slightly less rights" is going to be (understandably) rejected by A. The idea is not to find a middle ground on strong disagreements but to find nuggets of agreements in their views from which conversations can started. Statements like "Police should obey the law of the country" is maybe not going to be enthusiastically endorsed by A and B but is a possible ground for agreement.

One of the most positive effect is that both groups can be genuinely surprised by some of the other group opinions. B may not realize that A actually agrees on some anti-smuggling measures and A may not realize that B actually strongly approves of preserving native American rights. Reasons may diverge, implementations diverge, but fishing for agreements is a precious tool in order to mend societies.