Interactional Motivation in artificial systems: Between extrinsic and intrinsic motivation

Cognitive Systems Research

Riegler²

2019

Self Cite

Constitutive autonomy is the capacity of an entity to perpetually develop its individual constitution and coupling with its environment. We argue that computational entities (i.e., entities that can perform computation) can gain constitutive autonomy through motorsensory self-programminga mechanism by which the entity acquires new computational processes as a series of patterns of interaction that the entity can learn through experience, simulate internally, and enact in the environment. Motorsensory self-programming allows the evolution of the cognitive coupling between the entity's behavior selection mechanism and the environment as it appears from the viewpoint of the behavior selection mechanism. Constitutive autonomy of computational entities could lead to genuine agency.

Section: Selection Processmentioning

confidence: 99%

CASH only: Constitutive autonomy through motorsensory self-programming

Cognitive Systems Research

Riegler²

2019

Self Cite

“…We found that the EMDP model allowed us to encode metaphorical survival preferences by associating a value with the interactions that we define as concerning the agent's survival needs. These constraints generate a form of motivation that we call interactional motivation (Georgeon, Marshall, & Gay, 2012): the motivation to enact interactions with predefined positive values and to avoid interactions with predefined negative values. We predefine a slightly negative value for interactions that do not directly concern the agent's survival needs, to represent a light cost of enacting them.…”

Section: Self-motivationmentioning

confidence: 99%

ECA: An enactivist cognitive architecture based on sensorimotor modeling

Biologically Inspired Cognitive Architectures

Marshall

Manzotti

2013

Self Cite

A novel way to model an agent interacting with an environment is introduced, called an Enactive Markov Decision Process (EMDP). An EMDP keeps perception and action embedded within sensorimotor schemes rather than dissociated, in compliance with theories of embodied cognition. Rather than seeking a goal associated with a reward, as in reinforcement learning, an EMDP agent learns to master the sensorimotor contingencies offered by its coupling with the environment. In doing so, the agent exhibits a form of intrinsic motivation related to the autotelic principle (Steels, 2004), and a value system attached to interactions called interactional motivation. This modeling approach allows the design of agents capable of autonomous self-programming, which provides rudimentary constitutive autonomy-a property that theoreticians of enaction consider necessary for autonomous sense-making (e.g., Froese & Ziemke, 2009). A cognitive architecture is presented that allows the agent to autonomously discover, memorize, and exploit spatio-sequential regularities of interaction, called Enactive Cognitive Architecture (ECA). In our experiments, behavioral analysis shows that ECA agents develop active perception and begin to construct their own ontological perspective on the environment.

“…The EMDP model allows defining two forms of self-motivation: the motivation to be in control of one's own activity by seeking to successfully enact interactions, and the motivation to enact interactions that have predefined positive values and to avoid enacting interactions that have predefined negative values. We call the former autotelic motivation in reference to Steel's autotelic principle [10], and the latter interactional motivation [11].…”

Section: Enactive Markov Decision Process (Emdp)mentioning

confidence: 99%

An Enactive approach to autonomous agent and robot learning

2013 IEEE Third Joint International Conference on Development and Learning and Epigenetic Robotics (ICDL)

Wolf

2013

Self Cite

A novel way to model an agent interacting with an environment is introduced, called an Enactive Markov Decision Process (EMDP). An EMDP keeps perception and action embedded within sensorimotor schemes rather than dissociated. Instead of seeking a goal associated with a reward, as in reinforcement learning, an EMDP agent is driven by two forms of self-motivation: successfully enacting sequences of interactions (autotelic motivation), and preferably enacting interactions that have predefined positive values (interactional motivation). An EMDP learning algorithm is presented. Results show that the agent develops a rudimentary form of self-programming, along with active perception as it learns to master the sensorimotor contingencies afforded by its coupling with the environment.