Collective problem solving is supposed to benefit from cognitive diversity (e.g., when a team consists of individuals with different learning strategies). However, recent evidence for this claim fails to rule out an alternative explanation: that the benefit is due to moderate non-conformity, not diversity. We extend a previous agent-based simulation to distinguish these hypotheses, and demonstrate that diverse learning strategies alone do not yield the expected benefit. We extend the model further, based on an idea from the philosophy of science: Group-level benefits in complex problem solving often entail individual-level failures. Accordingly, we parameterize tolerance for failure, and show that there is an interaction between tolerance for failure and diversity. When tolerance for failure is zero, heterogeneous and homogeneous groups perform equally; when non-zero, diverse groups can outperform heterogeneous groups. Our agent-based simulations help clarify when cognitive diversity benefits collective problem solving.

We frequently use social information when making decisions.For instance, other people may know more about a problemthan we do, so we might update our initial beliefs in light oftheir opinions. The epistemic value of these social cues de-pends in part on their informational independence. Peopleshould thus be sensitive to nonindependence in their weightingof social information. However, the current literature yieldsconflicting results. In one recent study, participants valued so-cial information less when it was nonindependent; in another,participants were insensitive to nonindependence. We identifypossible causes of this inconsistency, and present an experi-mental paradigm that aims to fill these gaps. Then, in a study(N=200) with pre-registered hypotheses and analyses, we findthat participants were not sensitive to cue dependence. Wehighlight the relevance of this finding for the modern mediacontext, where nonindependence of both traditional and socialmedia sources can lead to the spread of bias or false belief.

Background

People with Parkinson's disease (PD) who develop visuo-perceptual deficits are at higher risk of dementia, but we lack tests that detect subtle visuo-perceptual deficits and can be performed by untrained personnel. Hallucinations are associated with cognitive impairment and typically involve perception of complex objects. Changes in object perception may therefore be a sensitive marker of visuo-perceptual deficits in PD.

Objective

We developed an online platform to test visuo-perceptual function. We hypothesised that (1) visuo-perceptual deficits in PD could be detected using online tests, (2) object perception would be preferentially affected, and (3) these deficits would be caused by changes in perception rather than response bias.

Methods

We assessed 91 people with PD and 275 controls. Performance was compared using classical frequentist statistics. We then fitted a hierarchical Bayesian signal detection theory model to a subset of tasks.

Results

People with PD were worse than controls at object recognition, showing no deficits in other visuo-perceptual tests. Specifically, they were worse at identifying skewed images (P < .0001); at detecting hidden objects (P = .0039); at identifying objects in peripheral vision (P < .0001); and at detecting biological motion (P = .0065). In contrast, people with PD were not worse at mental rotation or subjective size perception. Using signal detection modelling, we found this effect was driven by change in perceptual sensitivity rather than response bias.

Conclusions

Online tests can detect visuo-perceptual deficits in people with PD, with object recognition particularly affected. Ultimately, visuo-perceptual tests may be developed to identify at-risk patients for clinical trials to slow PD dementia. © 2018 The Authors. Movement Disorders published by Wiley Periodicals, Inc. on behalf of International Parkinson and Movement Disorder Society.