Civility and battling entrenched bias

Main article: Ratings system

Main article: The ratings system, human psychology and social dynamics

Entrenched bias is bias held by individuals that has persisted for a long time, perhaps from childhood. It is often unconscious and expressed in an automatic way. There doesn't appear to be a straightforward technical feature to combat it, other than what we are already contemplating (ie the ratings system). We clearly need to get people to think critically about their own beliefs. More precisely, how do we get people to believe the truth when they believe a falsehood and have done so for a long time?

Arguing usually does not work but empathetic listening does. Using evidence and facts helps but only if done in a non-confrontational way. The NYTimes columnist David Brooks argues, to give a recent example, that a growing disrespect for the working class explains much of their shift toward Trump. In his view, the Democrats are increasingly the party of the educated and have, as a result, started looking and feeling quite different than the working class. They try to use logic and facts to make their case. Biden’s attempt to reach them through economic policies is not having the desired effect because respect is more about feelings and attitudes than about objective benefits. It is notable here that the Democratic party is the one that believes in inclusiveness but seems to have no problem excluding those who don’t subscribe fully to its own notions of inclusiveness, even at the risk of electoral defeat.

It is hard to do this in person and hard to create an environment within an information system that achieves this. One way we could try to do it is to provide a rating for civility and filter out uncivil discourse on the platform. Most people will probably aim for civility and when they don’t, they are choosing not to. Their rating will reflect that choice. The system should probably be built by default with a fairly robust civility setting.

The ratings system, in addition, will rate highly those who are open minded and capable of interacting with a diverse set of people, whether they agree with them or not. Often, just keeping a dialogue open is key to keeping ideological discourse from spiraling out of control. The participants know there are real human beings on the other side of the debate and, most of the time, and recognize them as good people (even if they are wrong about something).

One specific technique, called deep canvassing, has been used to change voters’ minds and relies primarily on empathy as a tool. This article notes that

The political scientists Joshua Kalla and David Broockman have studied what campaigns can do to change voters’ minds during general elections, and the answer is basically nothing works.

The article goes on to describe deep canvassing as the method of choice in situations like these. However, the problem with deep canvassing is that it’s very expensive. Canvassers need to establish a one-on-one relationship with someone for maybe 15 minutes at a time and perhaps going back to them again a few more times for follow-up. This is clearly not something you can do on a large scale but is done with carefully targeted persuadable voters.

But, within these constraints, the method appears to be effective:

The results: deep canvassing changed attitudes by an average of around 4 to 6 percent, while traditional canvassing had little effect.

Deep canvassing might also help with a problem discussed by Krugman recently, that of “rural rage”, where rural voters prefer Trump to Biden even though Biden’s policies, according to Krugman, are more likely to help them economically.

If deep canvassing is indeed effective, how would we get our ratings system to “deep canvas” as part of an argument strategy? Well, for one thing, people can choose debate with those rated highly in empathy. Just having an empathy rating will make people more likely to want to score highly in this area. But another idea is to use AI to deep canvas. This is a creative endeavor which AI, surprisingly, seems to be fairly good at. Just tell it to pretend to be an empathetic debater on some topic and have a debate. The results are quite good. So for the “debate” feature of our system, we could start with AI setting a tone of civility which will help others follow along.

Another idea in boosting civility is to have a lag on posted material that gives authors a chance to think about what they have posted and make modifications. It especially gives them a chance to retract posts made in anger which they would later regret. This has been floated for social media and some forums, such as Kialo, effectively have it because posts must be approved by a human before publication. However, it would appear that major social media sites have not implemented this idea.

Whatever our strategy for promoting civility we should keep in mind that one of the central aims of our system is to get a controversial but correct opinion to the fore. Our system should enable this, not run counter to it as seems so often the case on social media. We have to create an environment where people can answer in good faith (not trampled on by negative ratings) and then be accepted by the community based on logic, veracity, etc. We’ve discussed rating the rater. Perhaps a simple rater objectivity index could be formulated so that, over time, we can identify those raters worth filtering for.