How Do Artificial Neural Networks Classify Musical Triads? A Case Study in Eluding Bonini's Paradox

How might artificial neural networks (ANNs) inform cognitive science? Often cognitive scientists use ANNs but do not examine their internal structures. In this paper, we use ANNs to explore how cognition might represent musical properties. We train ANNs to classify musical chords, and we interpret network structure to determine what representations ANNs discover and use. We find connection weights between input units and hidden units can be described using Fourier phase spaces, a representation studied in musical set theory. We find the total signal coming through these weighted connection weights is a measure of the similarity between two Fourier structures: the structure of the hidden unit's weights and the structure of the stimulus. This is surprising because neither of these Fourier structures is computed by the hidden unit. We then show how output units use such similarity measures to classify chords. However, we also find different types of units—units that use different activation functions—use this similarity measure very differently. This result, combined with other findings, indicates that while our networks are related to the Fourier analysis of musical sets, they do not perform Fourier analyses of the kind usually described in musical set theory. Our results show Fourier representations of music are not limited to musical set theory. Our results also suggest how cognitive psychologists might explore Fourier representations in musical cognition. Critically, such theoretical and empirical implications require researchers to understand how network structure converts stimuli into responses.     

Does integral affect influence intentions to use artificial intelligence for skin cancer screening? A test of the affect heuristic

Objective: This study investigated how affect influences people’s processing of messages about risks and benefits of using autonomous artificial intelligence (AI) technology to screen for skin cancer. We examined integral affect (emotion derived during decision making) separately from incidental affect (extraneous mood states).

Design: Using the affect heuristic framework, we randomly assigned 273 participants to conditions featuring risk (high, low, uncertain) or benefit (high, low, uncertain) messages about AI. Following ‘affect-as-spotlight’, we also explored whether people’s integral affect towards skin cancer moderated the relationship between risk/benefit messages and AI screening intentions.

Outcomes: Perceived risk, perceived benefit, positive and negative affect toward AI, intention to use AI screening.

Results: After controlling for incidental affect and risk perceptions, we found that compared to low risk messages, uncertain risk messages increased participants’ negative affect toward AI, decreased positive affect toward AI, increased AI risk evaluations and reduced AI benefit evaluations. Perceptual variables significantly mediated participants’ intentions to use AI for risk messages but not benefit messages. No moderation effects were found.

Conclusions: Results suggest extending the affect heuristic framework to include uncertain risk conditions. Integral AI affect influenced people’s interpretation of messages, which then impacted likelihood to use AI technology for health.