In the movie Marjorie | Prime (August 2017), John Hamm plays an artificial intelligence version of Marjorie’s deceased husband, visible to Marjorie as a holographic projection in her beachfront home. As Marjorie (played by Lois Smith) interacts with Hamm’s Prime through a series of one-on-one conversations, the AI improves its cognition by observing and processing Marjorie’s emotional expressions, movements, and speech. The AI also learns from interactions with Marjorie’s son-in-law (Tim Robbins) and daughter (Geena Davis), as they recount highly personal and painful episodes of their lives. Through these interactions, Prime ends up possessing a collective knowledge greater and more personal and intimate than Marjorie’s original husband ever had.
Although not directly explored in the movie’s arc, the futuristic story touches on an important present-day debate about the fate of private personal data being uploaded to commercial and government AI systems, data that theoretically could persist in a memory device long after the end of the human lives from which the data originated, for as long as its owner chooses to keep it. It also raises questions about the fate of knowledge collected by other technologies perceiving other people’s lives, and to what extent these percepts, combined with people’s demographic, psychographic, and behavioristic characteristics, would be used to create sharply detailed personality profiles that companies and governments might abuse.
These are not entirely hypothetical issues to be addressed years down the road. Companies today provide the ability to create digital doppelgangers, or human digital twins, using AI technologies. And collecting personal information from people on a daily basis as they interact with digital assistants and other connected devices is not new. But as Marjorie|Prime and several non-cinematic AI technologies available today illustrate, AI systems allow the companies who build them unprecedented means for receiving, processing, storing, and taking actions based on some of the most personal information about people, including information about their present, past, and trending or future emotional states, which marketers for years have been suggesting are the keys to optimizing advertising content.
Congress recently acknowledged that “AI technologies are rapidly evolving in capability and application throughout society,” but the US currently has no federal policy towards AI and no part of the federal government has ownership of the advancement of AI technologies. Left unchecked in an unregulated market, as is largely the case today, AI technological advancements may trend in a direction that may be inconsistent with collective values and goals.
Identifying individual rights
One of the first questions those tasked with developing laws, regulations, and policies directed toward AI should ask is, what are the basic individual rights–rights that arise in the course of people interacting with AI technologies–that should be recognized? Answering that question will be key to ensuring that enacted laws and promulgated regulations achieve one of Congress’s recently stated goals: ensuring AI technologies benefit society. Answering that question now will be key to ensuring that policymakers have the necessary foundation in front of them and will not be unduly swayed by influential stakeholders as they take up the task of deciding how and/or when to regulate AI technologies.
Identify individual rights leads to their recognition, which leads to basic legal protections, whether in the form of legislation or regulation, or, initially, as common law from judges deciding if and how to remedy a harm to a person or property caused by an AI system. Fortunately, identifying individual rights is not a formidable task. The belief that people have a right to be let alone in their private lives, for example, established the basic premise for privacy laws in the US. Those same concerns about intrusion into personal lives ought to be among the first considerations by those tasked with formulating and developing AI legislation and regulations. The notion that people have a right to be let alone has led to the identification of other individual rights that could protect people in their interactions with AI systems. These include the right of transparency and explanation, the right of audit (with the objective to reveal bias, discrimination, and content filtering, and thus maintain accountability), the right to know when you are dealing with an AI system and not a human, and the right to be forgotten (that is, mandatory deletion of one’s personal data), among others.
Addressing individual rights, however, may not persuade everyone to trust AI systems, especially when AI creators cannot explain precisely the basis for certain actions taken by trained AI technologies. People want to trust that owners and developers of AI systems that use private personal data will employ the best safeguards to protect that data. Trust, but verify, may need to play a role in policy-making efforts even if policies appear to comprehensively address individual rights. Trust might be addressed by imposing specific reporting and disclosure requirements, such as those suggested by federal lawmakers in pending federal autonomous driving legislation.
In the end, however, laws and regulations developed with privacy and other individual rights in mind, that address data security and other concerns people have about trusting their data to AI companies, will invariably include gaps, omissions, and incomplete definitions. The result may be unregulated commercial AI systems, and AI businesses finding workarounds. In such instances, people may have limited options other than to fully opt out, or accept that individual AI technology developers’ work was motivated by ethical considerations and a desire to make something that benefits society. The pressure within many tech companies and startups to push new products out to the world every day, however, could make prioritizing ethical considerations a challenge. Many organizations focused on AI technologies, some of which are listed below, are working to make sure that doesn’t happen.
Rights, trust, and ethical considerations in commercial endeavors can get overshadowed by financial interests and the subjective interests and tastes of individuals. It doesn’t help that companies and policymakers may also feel that winning the race for AI dominance is a factor to be considered (which is not to say that such a consideration is antithetical to protecting individual rights). This underscores the need for thoughtful analysis, sooner rather than later, of the need for laws and regulations directed toward AI technologies.