But the thing is, they specifically don't isolate 'social stimulus'. They offer new information, and those who knew they were guessing revised their estimates.

But the thing is, they specifically don't isolate 'social stimulus'. They offer new information, and those who knew they were guessing revised their estimates. Those who had a good idea of the value did not. There is nothing 'social' here at all. A better control would be to ask them how sure they were of their answers, and use that to adjust the bias.