Specifically for updating, I can imagine a test where the person is gradually given more and more information; the initial information is an evidence of an outcome “A”, but most of the latter information is an evidence of an outcome “B”. The person is informally asked to make a guess soon after the beginning (when the reasonable answer is “A”), and at the end they are asked to provide a final answer. Some people would probably get stuck as “A”, and some would update to “B”. But the test would involve some small numbers, shapes, coins, etc.; not real-life examples.
I’ve seen experiments that tested this; I thought they were mentioned in Thinking and Deciding or Thinking Fast and Slow, but I didn’t see it in a quick check of either of those. If I recall the experimental setup correctly (I doubt I got the numbers right), they began with a sequence that was 80% red and 20% blue, which switched to being 80% blue and 20% red after n draws. The subjects’ estimate that the next draw would be red stayed above 50% for significantly longer than n draws from the second distribution, and some took until 2n or 3n draws from the second distribution to assign 50% chance to each, at which point almost two thirds of the examples they had seen were blue!
I’ve seen experiments that tested this; I thought they were mentioned in Thinking and Deciding or Thinking Fast and Slow, but I didn’t see it in a quick check of either of those. If I recall the experimental setup correctly (I doubt I got the numbers right), they began with a sequence that was 80% red and 20% blue, which switched to being 80% blue and 20% red after n draws. The subjects’ estimate that the next draw would be red stayed above 50% for significantly longer than n draws from the second distribution, and some took until 2n or 3n draws from the second distribution to assign 50% chance to each, at which point almost two thirds of the examples they had seen were blue!