Report

Teaser, summary, work performed and final results

Periodic Reporting for period 2 - METAWARE (Behavioral and neural determinants of metacognition and self-awareness in human adults and infants)

Teaser

One of the most fundamental issues in contemporary science concerns the nature of consciousness. The project METAWARE explores the neural determinants of conscious and unconscious cognitive processes, by investigating them during sleep, self-recognition and metacognition. More...

Summary

One of the most fundamental issues in contemporary science concerns the nature of consciousness. The project METAWARE explores the neural determinants of conscious and unconscious cognitive processes, by investigating them during sleep, self-recognition and metacognition. More specifically, the objectives of the project are the following follow:

- A first objective of the project consists in providing news insights about unconscious cognitive processing in the full absence of consciousness, including self-consciousness and metacognitive abilities, by focusing on the sleeping brain and its ability to process and learn information from its surrounding environement.

- A second objective of the project is to explore the fundamental issue of whether multiple agents can share information and each other’s conscious access mechanisms, without being aware of it.

- A third objective of the project is to answer the two fundamental issues of whether infants have a capacity for metacognition (do they know they know) and whether they experience self-consciousness (do they feel themselves as a unitary entity).

This project is intrinsically interdisciplinary, addressing long-standing issues in philosophy and mind sciences (e.g., Why do we sleep? Is consciousness private to the individual? Can we share minds? Are infants aware about themselves? Should they be considered as ‘real persons’ from an ethical point of view?) by using both behavioral paradigms originated from experimental psychology and brain imaging methods from neuroscience such as fMRI and EEG. The project has also medical implications as it might help paediatricians confronting issues of infant consciousness in relation to anaesthesia, pain, and pathologies. In addition, this project relies on the most recent advances in engineering and machine learning, by using brain-computer interface methods with adults and by using augmented reality setups to explore how the self develops in infants.

Work performed

The Metaware project is carried out without major difficulties and has already been successful in leading to several high-impact publications in either cognitive psychology/neuroscience or general science. During this first half period, the project already produced several breakthrough results leading to high-rank publications. Some of this research also received large media coverage (Time Magazine, Washington Post, Le Monde, The Scientist, and many others) for 3 papers in particular, one published in Science and one published in PNAS (Proceedings of the National Academy of Science) for WP3, and the other one in Nature communications in 2017 for WP1. Below we enumerate all the work packages of the project and their sub-components, while describing their level of progress, the results achieved so far and the corresponding publications.

Our work on the physiological markers of information processing and learning during sleep revealed that the human brain continues to respond to external stimulation and can even be induced to generate classifications of incoming sounds. We show that this is the case in particular in light sleep, but that responsiveness to the external world disappears during deep sleep stages (Andrillon et al., 2016, Journal of Neuroscience). Using a We also showed that, again during light but not during deep sleep, the human brain can continue responding and even adapting to (i.e., learning from) the acoustic environment (Andrillon & Kouider, 2016, Neuroscience of Consciousness; Andrillon et al., 2017, Nature communications). Furthermore, another study from our group reveals that sleeper can selectively attend to relevant information in the acoustic environment (when several people are talking at the same time, as in the so-called “cocktail party” situation, the sleeper’s brain selectively amplifies the speech stream which is most relevant to her/him; Legendre et al., in press). Congruent with our previous studies, this research shows that selective attention is possible during light but not during deep sleep.

- Legendre, G., Andrillon, T., Koroma, M., & Kouider S. (in press). Attentional tracking of relevant signals during human sleep. Forthcoming in Nature Human Behaviour.
- Andrillon, T., Pressnitzer, D., Léger, D., & Kouider S. (2017). Formation and suppression of acoustic memories in human sleep. Nature communications, 8 (1), 179.
- Andrillon, T., Poulsen, L.K., Hansen, L.K., Léger, D., & Kouider, S., (2016). Neural Markers of Responsiveness to the Environment in Human Sleep. The Journal of Neuroscience, 36 (24), 6583-6596.
- Andrillon, T., & Kouider, S. (2016). Implicit Memory for Words Heard During Sleep. Neuroscience of Consciousness, 25, 2823-282.

Our work on the precursors of metacognition and self-consciousness in infants used measures of implicit behaviors and neural signatures to reveal that infants possess ‘core’ metacognitive abilities. Indeed, they can communicate in a non-verbal manner whether they know they don’t know (Goupil, Romand-Monnier & Kouider., 2016, Proceeding of the National Academy of Science - PNAS). Moreover, they evaluate their decision confidence and generate electrophysiological signals (i.e., the Error-Related Negativity brain wave) reflecting the internal monitoring of their own errors after having made a simple decision (Goupil & Kouider, 2016; Current Biology). Another manuscript is currently under review where we propose a new theoretical framework for the development of metacognition and self-reflection in infants. In addition, our work provides new insights from what we know now about consciousness and metacognition in babies to the recent efforts in developing conscious artificial intelligence using deep learning algorithms (Dehaene, Lau & Kouider, 2017).

- Dehaene, S., Lau, H. & Kouider, S. (2017). What is consciousness, and could machines have it? Science, 358(6362), 486-492.
- Goupil, L., Romand-Monnier, M., & Kouider S. (2016). Infants ask for he

Final results

For the rest of the project, we expect to provide new insights on two aspects of the project for which we have already put a lot of efforts in building the technology necessary to address these issues during the first half period, but for which a large amount of work remains to be done now with experimentations during the remaining half of the project.

A first aspects concerns the development of the infant self beyond the traditional mirror test, in order to provide new insights on the existence of primitive forms of self-consciousness in infants. The only existing behavioral index of self-consciousness in children (and non-human primates) is the mirror test: does the child recognize the mark on his face as his own when facing a mirror. Normally the mirror test is not passed before the end of the second year of life (18-24 months). In the past few years, we investigated a new approach relying on augmented reality that we believe will allow us demonstrating that younger infants (i.e., below 1 year of age) already have a sense of themselves.

The second aspect concerns the work on sharing and confounding conscious access mechanisms without knowing. This project required large efforts and resources in technical engineering to integrate real-time brain-computer interfaces, machine learning algorithms, psychophysics and virtual reality environments. During the first half period, our ability to build a reliable non-invasive brain-computer interface has overpassed our expectations. Indeed, we have been able to build over the past few years the most robust and fastest machine learning algorithm for recording EEG data, decoding and interpreting its cognitive dimensions and transform it into a command over a digital interface, all in less than half a second for certain tasks. This allow subjects to selects contents and perform actions on a computer directly with their mind almost instantaneously. By having several people performing collective actions over a computer directly with their brain, we expect to provide new insights on how humans can share their minds in a purely neural manner (i.e., without motor or verbal information). We also expect to provide new insights on whether people the performance of people connecting their brains for a given task will, as whole, be better than the sum of the individuals.

And of course, we should also expect unexpected findings, which is the best way to scientific discoveries !