Jean-Louis Dessalles

Original theoretical results

Here is a list of original theoretical results that I believe to be of some interest.

Simplicity Theory

Interesting events are simpler to describe than to generate.
This law predicts that events close in time and space are more likely to raise interest.
It also predicts the fascination of coincidences, story rounds, and several other spectacular aspects of spontaneous communication.

Generative theory of spontaneous dialogue

For long I suspected that the way people conceive their next move in a discussion is quite "mechanical". During years I strived to simplify a cognitive algorithm to bridge the gap between knowledge and conversation. I arrived at a minimal procedure, named CAN (Conflict, Abduction, Negation) that correctly predicts important aspects of argument generation in spontaneous discussion.

Altruism, social signalling, and the evolutionary emergence of language

Human language remains a mystery for evolutionary theory: why give information to competitors?
My answer: Language evolved to display social signals.
This evolution was possible in the particular political context of our species.
My model of Social Signalling also provides a new explanation for many cases of apparently altruistic behaviour, in several species, independently from language.

The nature of semantic knowledge

What is the nature of meaning? What is the nature of the representations we attached to "apple", "cut", "cut an apple"?
We could show that any account of meaning using permanent symbolic representations is inconsistent. On the other hand, non-symbolic models do not account for systematic properties of language and reasoning. Together with Laleh Ghadakpour, we designed a model in which meaning is dynamically produced by cognitive operators like the contrast operator.

A cognitive model of Aspect

All human individuals have an intuitive grasp of Aspect. "She smoked during the lecture" could mean a single short smoking event (included in the lecture event), a single long event (covering the lecture event), or a repeated event. All attempts to capture this intuitive competence into a simple algorithme have failed.
Together with Laleh Ghadakpour and Damien Munch, we showed that models in which Time is implemented in a static structure are inconsistent. We could capture an important part of the Aspectual competence as a minimal recursive procedure, by using new notions such as "predication".

Problem solving isn’t purely procedural.

Together with Jean-Bernard Auriol, we showed that human problem solving involves both procedural and logical abilities. The challenge was to explain how both co-operate.

Other important contributions

Non epiphenomenality of qualia

Thanks to an evolutionary argument, I could show that epiphenomenal accounts of qualia are inconsistent. Qualia are (locally) optimal, as any feature involved in the "Darwinian loop". Therefore, they can’t be epiphenomenal (but the ‘explanatory gap’ remains unfortunately intact).


What matters isn’t innateness, but anisotropic bias. I show that the nature-nurture debate should be rephrased: rather than invoking some ill-defined "quantity" of innate knowledge, it should oppose isotropic to anisotropic biases.

Emergence = complexity drop.

Together with Eric Bonabeau, we showed how all cases where emergence is invoked correspond to a complexity drop.

(other topics I investigated: see books and papers).