Excellent analysis! That 'belief-consistent information procesing' idea is such an elegant way to frame things. It really makes you wonder how many of our deeply held 'priors' are just stubborn defaults. It's so hard to debug those mental algorithms sometimes, even when we try. Makes me think of bias in AI training, too!
Yes, it’s interesting to consider if AI also employs belief consistent information processing – I suspect the answer is yes as it’s trained on human reasoning traces, but perhaps one can make an AI more truth-seeking through a combination of prompting & RLHF?
One of the other main aspect this framing emphasizes IMO is the importance of having a diversity of inputs. This is somewhat obvious in an age of social media & echo chambers, but sourcing such views is difficult. Reading older books & papers is one approach I’ve seen others take.
Excellent analysis! That 'belief-consistent information procesing' idea is such an elegant way to frame things. It really makes you wonder how many of our deeply held 'priors' are just stubborn defaults. It's so hard to debug those mental algorithms sometimes, even when we try. Makes me think of bias in AI training, too!
Yes, it’s interesting to consider if AI also employs belief consistent information processing – I suspect the answer is yes as it’s trained on human reasoning traces, but perhaps one can make an AI more truth-seeking through a combination of prompting & RLHF?
One of the other main aspect this framing emphasizes IMO is the importance of having a diversity of inputs. This is somewhat obvious in an age of social media & echo chambers, but sourcing such views is difficult. Reading older books & papers is one approach I’ve seen others take.