Theories of consciousness have a long and controversial history. One well-known proposal — integrated information theory — has recently been labeled as ‘pseudoscience’, which has caused a heated open debate. Here we discuss the case and argue that the theory is indeed unscientific because its core claims are untestable even in principle.

Understanding the psychological computations underlying motivation can shed light onto motivational constructs as emergent phenomena. According to Murayama and Jach, reward-learning is a key candidate mechanism. However, there's no such thing as a free lunch: Not only benefits (like reward), but also costs inherent to motivated behaviors (like effort, or uncertainty) are an essential part of the picture.

The posterior parietal cortex and frontal eye field contain maps of visual salience on which the decision to choose a saccade may be based. However, an averaging express saccade is not represented by a victorious unimodal representation in the superior colliculus. Normalization as described by Findlay & Walker is not necessary for the generation of saccades.

We argue that the type of meta-learning proposed by Binz et al. generates models with low interpretability and falsifiability that have limited usefulness for neuroscience research. An alternative approach to meta-learning based on hyperparameter optimization obviates these concerns and can generate empirically testable hypotheses of biological computations.