The 95%-CI decision rule in psychological assessment from a decision theoretic perspective Authors: Florian Pargent, David Goretzko, Benedikt Friemelt Psychological assessment sometimes requires a dichotomous decision, whether a person is “below the norm” in some psychological domain. One common decision rule is whether the upper boundary of a 95% confidence interval for the person’s true value lies below the 16% quantile (one standard deviation below the mean of a normal distribution) based on the distribution of the true values in the population. We demonstrate how the 95%-CI decision rule can be interpreted from the perspective of (Bayesian) decision theory: it is equivalent with minimizing expected loss when considering the type I error as 39 times more severe than the type II error. Many practitioners might not be aware of this implicit assumption. Thus, mindlessly applying any default rule can lead to diagnostical decisions which are not in line with the diagnostician’s judgement of the possible negative consequences of decision errors in the concrete setting. In fact, “rational” optimal decisions would require an explicit weighting of those errors. To stress this point, we present a small survey of clinical neuropsychologists, who had to report different representations of their internal loss function for a fictitious diagnostical scenario. We look at how much these judgements differ between practitioners and how much decisions based on practitioners’ “true” loss functions differ from the 95%-CI rule.