Background/field: James’s doctorate was in English literature. He then cross-trained in experimental psychology followed by computational linguistics – and is now combining all three in a project on how mental illness interacts with textual structures.

Current research: Using artificial intelligence / machine learning to investigate the therapeutic relevance of literature, with a focus on anxiety disorders and depression.

Research questions and measures: How can the level of entropy (roughly equivalent to unpredictability) in a text be quantified, and how does it affects readers’ responses?


Here’s James’s summary of his take on entropy and cognition, and how entropy is manifested in language.

A good way to think of entropy is to imagine a biased coin, which turns up heads nine-tenths of the time: if you guess that any given toss will turn up heads, you’ll be correct most of the time—this system is highly predictable and thus has low entropy. An unbiased coin has high entropy, because guessing heads has no predictive advantage.

In language, entropy is usually used to measure how predictable the occurrence of a linguistic symbol is, relative to the sequence of symbols that go before it. Take the phrases ‘Northern Ireland’, ‘Northern lights’, and ‘Northern footpath’: the first two have low entropy, because the second word occurs with a high probability when the first is present; the third, by contrast, is not a common phrase in regular English, and thus has high entropy. This example is simple, but it can be readily extended to more complex forms of discourse.

The value of entropy is that it gives us a good measure of abstraction: low-entropy systems are detail-sparse and predictable; high entropy systems are detail rich and unpredictable. Moreover, there exists a branch of social psychology called construal level theory, which details the effects of abstraction on cognition. In particular, it shows how abstract (i.e. low-entropy) stimuli expand the mental horizons of the people experiencing them with respect to time, space, likelihood, and social distance, while concrete stimuli contract them along the same dimensions.

Hypotheses: Depression involves disengagement from everyday realities, so should be benefited by rich, concrete, higher-entropy stimuli to encourage re-engagement. Anxiety, by contrast, involves obsessive rumination on minute details, so should be benefited by more abstract, stripped-down, lower-entropy stimuli to lift readers out of here-and-now obsessions.


  • Because of the contrastive method, the need to exclude cases where anxiety and depression are comorbid.
  • The possibility that not all manifestations of depression and anxiety can be neatly categorised in terms of high/low entropy levels.

What’s next? Asking bigger-picture questions like: are high-entropy texts are more likely to be created in societies with high levels of predictability? In general, taking a cue from Lévi-Strauss, conceiving large-scale projects in which anthropology unfolds out of computation.

Other topics we touch on:

  1. The purpose of research: the concept of ‘usefulness’, the gap between claimed and genuine intentions (in literary studies), the similarities between literary studies and religion.
  2. The (not necessarily good) reasons why literary scholars and psychiatrists may be drawn to this kind of medical humanities research, and the possibilities for good to come out of these relationships of convenience nonetheless. The structural reasons why healthcare can benefit from the humanities: countering the medical tendency to a zoomed-in reduction of an entire context to a single manipulable factor.
  3. The (not necessarily good) reasons why literary scholars seek out literature.

Find out more:

  • James’s Brunel University webpage is
  • His 2018 paper in Review of General Psychology, ‘People searching for meaning in their lives find literature more engaging’, is here
  • An example of James’s earlier work on narrative and social cognition is his 2014 PLOS ONE paper, ‘Inference or enaction? The impact of genre on the narrative processing of other minds’, here


James’s comments on the image, Wyndham Lewis’s ‘A Battery Shelled’ (1919):

Lewis’s subject matter – trench warfare – evinces anxiety to the highest degree, to the extent that death or injury can supervene at any moment. Here, this anxiety is distanced using a low-entropy representational style, where naturalistic elements are subordinated to a predictable, quasi-geometrical scheme. As such, it exhibits a common impulse in abstract art: the neutralising of historically conditioned anxiety by way of an entropy-reducing aesthetic.


And thanks to Aussens@iter for the intro/outro music: 

Between Worlds (Instrumental) by Aussens@iter (c) copyright 2017.
Licensed under a Creative Commons Attribution (3.0) license. Ft: (Smiling Cynic)