Thread by @chrissanders88

An interesting study on the effects of prediction error on how people update their beliefs on topics.

Overview article: psypost.org/2021/11/psycho…

Research article: scholar.princeton.edu/sites/default/…

Relevance here to combatting misinformation.

The gist of the findings is that folks are more likely to change their mind on a topic when asked to make a prediction about some facts relevant to the topic and subsequently finding out their prediction was false.

Further, the magnitude of the prediction error is notable:

"we found that prediction error size linearly predicts rational belief update
and that making large prediction errors leads to larger belief updates than being
passively exposed to evidence"

As I understand, this isn't saying that exposing folks to prediction error will change their mind, more so that exposing them to prediction error actively, deliberately, and quickly has a greater chance to change minds than them passively finding the error on their own.

That lines up with the idea that one of the strongest techniques for getting folks to change their view on a topic is to get them to sincerely consider the alternatives. The surprise that comes along with prediction error is one way to do that.

So here, we're looking at a conversational technique to stimulate consideration of alternative views.

We know that folks who arrive at a conclusion absent logic are unlikely to change their mind based on logic. A big part of that is identity and the things that folks wrap up in it. It's hard to challenge folks based on things they consider part of their identity.

A lot of research in this area points not to tell people what to think, but to stimulate their thoughts in ways that connect other parts of their identity to alternative viewpoints. Of course, this gets used in positive and negative ways alike.

When you consider how you deliberately stimulate this kind of prediction error... here's an example.

Let's say you're discussing the merits of govt welfare programs with someone who is very against it and you want them to consider alternate perspectives.

You could ask something like, "What percentage of folks receiving SNAP (food stamp) benefits are receiving those payments fraudulently?"

Chances are, the person you are talking to will think it's quite high in this example.

Because fraud detection is imperfect the actual stat will vary, but estimates range from 3-7 percent. That is often a lot less than folks who are adamantly against such programs think absent real data. A lot of opportunity for prediction error here!

Not only that, this opens up some interesting conversations. "Wait, how do they know that?"

Many don't even realize that the govt or other groups know about or track welfare fraud in any meaningful way.

"Well, why don't they fix it?"

It's a trade-off, like most things. If you over-correct the system problems that allow the fraud, you might exclude people who legitimately need help. So, you accept some level of fraud to make sure you can help people that need it.

In reality, that's how a lot of government programs work. There is a trade-off between some level of allowable fraud while still allowing the program to serve those who need it.

We see this in business, too. Retail stores know some theft will happen. But to completely eliminate it would require some uncomfortable and draconian decisions that would send customers elsewhere. They accept some theft in exchange for a better customer experience.

When you think about the argument this way, it gives you another way to approach the person this scenario is set up around.

Start with this question:
"What percentage of SNAP payments being considered fraud would warrant scrapping the whole program?"

Now, the person has to consider what amount of fraud they think there is already along with their own subjective cut-off for the program being worth it to society.

"I think anything greater than 20% means it's not worth it."

Well, I've got some good news for you!

Now you've created an opportunity for a prediction error, valuable discussion, and a better understanding of the tradeoffs.

None of this is a panacea, it's just a technique that research identifies as helpful in combatting misinformation and changing minds. Of course, these things work best with people acting in good faith. So, more your family at Thanksgiving than random Facebook commenters.