Comments on: Gullible Then Skeptical

By: Steve Rivers

Steve Rivers — Thu, 26 Jul 2007 20:11:59 +0000

>why we seem too eager to believe the first opinion we hear on a
>subject, and then seem too skeptical about further contrary
>opinions we hear

Maybe we just find it unpleasant to be wrong, or rather to admit being wrong.
We might also see changing an opinion as a sign of weakness (i.e. we are less knowledgeable, impressionable, haven’t got our thoughts straight, etc.)

By: Jason Malloy

Jason Malloy — Tue, 24 Jul 2007 07:02:35 +0000

why we seem too eager to believe the first opinion we hear on a subject, and then seem too skeptical about further contrary opinions we hear.

I am skeptical of this claim (which is the first opinion I’ve heard on the subject). Reference?

By: LemmusLemmus

LemmusLemmus — Fri, 20 Jul 2007 20:41:23 +0000

We tend to view nonutterances as agreement or at least non-disagreement. Someone utters an opinion, nobody says anything against that view, the listener assumes that there is more weight to the opinion than just one person holding it.

The second view, uttered by a different person, has no such extra weight.

Prediction: If a number of people utter an opinion, and an equal number of people utter the opposite opinion immediately afterwards, the effect does not occur.

By: Hopefully Anonymous

Hopefully Anonymous — Fri, 20 Jul 2007 11:30:13 +0000

Another classic example of this phenomenon (or at least cultural belief that it exists) is the common emphasis on the importance of making good first impressions upon meeting someone.

By: Constant

Constant — Fri, 20 Jul 2007 02:36:08 +0000

Thanks Eliezer. That was what I suspected and hoped.

By: Brad

Brad — Fri, 20 Jul 2007 02:02:25 +0000

This was a very clever posting. I happen to believe that people tend to be more open to the second opinion they hear about something. See where I’m going with this? Now you all believe me, not Ute Shaw. Of course, had Robin posted that Ute believed we were partial to the second thing we heard, and I chimed in saying we preferred the first, then we’d have a real paradox.

Anyone who’s ever been a kid or been around kids knows these theories are suspect. Kids in trouble try to take advantage of this phenomenon (as they imagine it) by telling their side first. Adults quickly learn to wait until they have all sides before they decide the facts. Which, BTW, totally frustrates kids who try to get their story out first. Adults who act like kids in this respect tend to really piss other adults off.

By: Matthew C

Matthew C — Thu, 19 Jul 2007 22:32:55 +0000

Bayesian inference is not influenced by the order in which updates are applied. p(c|b,a) = p(c|a,b).

This assumes that there is no confirmation bias effect in how we evaluate additional evidence. Of course, in real life, confirmation bias, group sociological bias, endowment effects and all sorts of other biases come into play and make the order of our exposure to ideas very important.

By: Eliezer Yudkowsky

Eliezer Yudkowsky — Thu, 19 Jul 2007 22:26:07 +0000

Bayesian inference is not influenced by the order in which updates are applied. p(c|b,a) = p(c|a,b).

In theory, the order of observations can itself count as an observation. But once you collect your observations, probability theory takes no notice of the order in which you apply the updates.

By: Constant

Constant — Thu, 19 Jul 2007 19:23:00 +0000

“When we first hear something, we move our prior that way. This decreases the prior for the opposite being true, and we discount that information more easily.”

Yes, and this isn’t a bias unless we move our prior too much. That is, even while we receive the second evidence more skeptically, the end result may still be the same regardless of which evidence we received first. So, if there is a bias here, it is not that we receive the second evidence more skeptically, but that we do this to too great a degree. So diagnosing a bias would involve more than merely noticing that the second evidence is met with greater skepticism.

(I have not done the math, so I do not know whether repeated Bayesian inference is influenced by the order of observations, so I don’t know whether Bayesian inference is biased towards or against the first observation.)

By: Matthew C

Matthew C — Thu, 19 Jul 2007 18:12:26 +0000

When we first hear something, we move our prior that way. This decreases the prior for the opposite being true, and we discount that information more easily. Yep. On my blog, I discuss a very interesting study that shows just how biased we all tend to be when evaluating data that opposes our priors.