Even that .69%-acceptable statistic may be a political maneuver. I found a meta analysis a year or two ago of AI healthcare diagnostics that found about this level of acceptability in the literature.
Where it becomes political is that a prestigious doctor friend unsympathetic to AI diagnosis used this statistic to blow off the whole field, rather than to become interested in the tiny fraction of acceptable research. Which is political on its own, and also has to make you wonder if researchers set their quality bar to get the result they want.
Nevertheless it IS discouraging that about 276⁄40000 papers would be acceptable.
Even that .69%-acceptable statistic may be a political maneuver. I found a meta analysis a year or two ago of AI healthcare diagnostics that found about this level of acceptability in the literature.
Where it becomes political is that a prestigious doctor friend unsympathetic to AI diagnosis used this statistic to blow off the whole field, rather than to become interested in the tiny fraction of acceptable research. Which is political on its own, and also has to make you wonder if researchers set their quality bar to get the result they want.
Nevertheless it IS discouraging that about 276⁄40000 papers would be acceptable.