I have a design where I present participants with a series of decisions, and then ask them about some of these decisions later. Similar work has found that even when the particular decision was made in error, and people 'should' realise this, participants carry on and attempt to justify the decisions anyway. I presented similar decisions but with a computer program acting as 'the experimenter', providing text boxes for input. Participants mentioned errors made in my study much more - they didn't carry on justifying an error.
This was an unexpected result. I suspect some of this may be due to participants not wanting to admit to making an error when they feel more identifiable, but I don't know where to start in terms of literature on the topic!
So essentially - what are the differences in peoples willingness to admit and correct their own, or even others, errors, between social and computer interactions? and further, what causes these differences?