My intuition was that participants assumed the human forecasters could use some knowledge the model couldn’t, but that is definitely not the case in the xperiments. The model and the forecaster have exactly the same data and that is made clear in the setup.
It is really that participants prefer the human despite ‘better knowledge’ - esp. if the comparison is explicit:
Interestingly, participants in the model-and-human conditions, most of whom saw the model outperform the human in the first
stage of the experiment [...] were, across all studies, among those least likely to choose the model.
Page 6, emphasis mine.
I admit that I am surprised and I do not understand the cause of this algorithm aversion.
My intuition was that participants assumed the human forecasters could use some knowledge the model couldn’t, but that is definitely not the case in the xperiments. The model and the forecaster have exactly the same data and that is made clear in the setup. It is really that participants prefer the human despite ‘better knowledge’ - esp. if the comparison is explicit:
Page 6, emphasis mine.
I admit that I am surprised and I do not understand the cause of this algorithm aversion.