Table 1.

The five sets of simulations performed, each of n = 972 agents. Parameter values were selected pseudorandomly from the ranges shown.

Model	Figures	Frequency of delusions with thresholdsa of 66% (60%, 70%)	Parameters, their ranges, and (below) Spearman correlations with delusion scores
			Likelihood precision $\left[\begin{array}{cc}a& 1-a\\ 1-a& a\end{array}\right]$ a = 0.60 − 0.99	Habit resistance Dir(e) = 2 − 600	Initial consistency 1–125 trials	Mood [c − c] = −4.5 − 4.5	Choice precision α = 0.5 − 2.75	Policy precision $\frac{1}{\beta }=0.25-1.75$
No affect, consistently trustworthy sequence	2C, 5A	0% (0%, 0%)	ρ = −0.54	ρ = −0.40			ρ = 0.022	ρ = 0.32
With affect, consistently trustworthy sequence	4D, 5B, 6A	1.3% (1.7%, 1.3%)	ρ = −0.63	ρ = −0.20		ρ = −0.008	ρ = 0.017	ρ = 0.47
No affect, changing trustworthiness sequence	5C	1.0% (1.7%, 0.6%)	ρ = −0.44	ρ = −0.38	ρ = −0.034		ρ = 0.037	ρ = 0.34
With affect, changing trustworthiness sequence	4E, 4F, 5D, 6B, 7A	3.2% (4.6%, 2.0%)	ρ = −0.64	ρ = −0.20	ρ = −0.032	ρ = 0.039	ρ = 0.017	ρ = 0.48
With affect, changing trustworthiness sequence, with treatment	6C, 7B	0.5%	ρ = −0.63	ρ = −0.20	ρ = −0.036	ρ = 0.034	ρ = 0.020	ρ = 0.47

^aThresholds for falsity, certainty and incorrigibility criteria – see Section 2.3.7 – with results from alternative thresholds for all three criteria given in brackets.