Towards the Mind of a Humanoid: Does a Cognitive Robot Need a Self? - Lessons from Neuroscience

Elena Antonova, Chrystopher Nehaniv

Research output: Chapter in Book/Report/Conference proceedingConference contribution

58 Downloads (Pure)


As we endow cognitive robots with ever more human-like capacities, these have begun to resemble constituent aspects of the ‘self’ in humans (e.g., putative psychological constructs such as a narrative self, social self, somatic self and experiential self). Robot’s capacity for body-mapping and social learning in turn facilitate skill acquisition and development, extending cognitive architectures to include temporal horizon by using autobiographical memory (own experience) and inter-personal space by mapping the observations and predictions on the experience of others (biographic reconstruction). This ‘self-projection’ into the past and future as well as other’s mind can facilitate scaffolded development, social interaction and planning in humanoid robots.

This temporally extended horizon and social capacities newly and increasingly available to cognitive roboticists have analogues in the function of the Default Mode Network (DMN) known from human neuroscience, activity of which is associated with selfreferencing, including discursive narrative processes about present moment experience, ‘self-projection’ into past memories or future intentions, as well as the minds of others. Hyperactivity and overconnectivity of the DMN, as well as its co-activation with the brain networks related to affective and bodily states have been observed in different psychopathologies. Mindfulness practice, which entails reduction in narrative self-referential processing, has been shown to result in an attenuation of the DMN activity and its decoupling from other brain networks, resulting in more efficient brain dynamics, and associated gains in cognitive function and well-being. This suggests that there is a vast space of possibilities for orchestrating self-related processes in humanoids together with other cognitive activity, some less desirable or efficient than others. Just as for humans, relying on emergence and selforganization in humanoid scaffolded cognitive development might not always lead to the ‘healthiest’ and most efficient modes of cognitive dynamics. Rather, transient activations of self-related processes and their interplay dependent on and appropriate to the functional context may be better suited for the structuring of adaptive robot cognition and behaviour.
Original languageEnglish
Title of host publicationALIFE 2018
Subtitle of host publicationProceedings of the Artificial Life Conference 2018 (ALIFE 2018 - Tokyo, Japan)
EditorsTakashi Ikegami, Nathaniel Virgo, Olaf Witkowski, Mizuki Oka, Reiji Suzuki, Hiroyuki Iizuki
PublisherMIT Press
Number of pages419
Publication statusPublished - 18 Jul 2018


Dive into the research topics of 'Towards the Mind of a Humanoid: Does a Cognitive Robot Need a Self? - Lessons from Neuroscience'. Together they form a unique fingerprint.

Cite this