Consciousness as the Foundation of Legal Agency in AGI

Abstract

This paper argues that it is time to address the legal status of emerging Artificial General Intelligence (AGI) that plausibly replicates human behavior.  It concludes that a new juridical entity, “Legally Conscious Persons”, should be created for that category.  This status is based on the model of “legal personhood” which stipulates specific rights and duties for corporate entities, but has a potentially broader scope.

Crimes and intentional torts require “mens rea” which implies a conscious, rational agent having an intention to take (or not take) an act.  Traditionally, living natural persons are thought to be the only category which fits this requirement.  To suggest attributing consciousness to a non-organic artifact (e.g., AGI) creates a legal conundrum: If AGI is not alive, it cannot be conscious.  If AGI is conscious, what does that say about the nature of life?  Would a conscious AGI be a person, a thing, or something else?

The paper reviews the origins of life and the coevolution through stages of the human brain and consciousness.  It then turns to the contested issue of AGI consciousness.  It reviews the scholarly and expert debate over the nature and source of consciousness and concludes that while there is no consensus on its nature, there are many theories which can provide indicators of its presence.  It reviews these theories through the lenses of ontology, epistemology and axiology.  It then introduces the proposed technical and behavioral indicators of the likelihood of consciousness.

Human consciousness is not well understood.  Neurological indicators correlate with but are not the same as “qualia”, a person’s subjective experience.  Likewise with AGI, “consciousness”, if it exists, can only be predicted by technological and behavioral indicators.  For purposes of recognizing an AGI as “legally conscious” it must show indicators that demonstrate a high likelihood of meeting the criteria for “legally conscious”.  A process for reaching these criteria based on existing models is offered.

The time to begin to establish these indicators is now.  The Paper offers a series of specific recommendations for government, enterprise and civil society to meet this challenge.  Failure to do so, it concludes, would someday be seen as the equivalent of malpractice.

Keywords: AGI/Artificial General Intelligence, Consciousness, Juridical Person, Sentient AGI, Legally Conscious, Thinking Machines