In addition to the expo floors spread across different venues (I clocked 8.5 walking miles on Wednesday), CES features an immense range of conference sessions and collocated meetings. I attended several sessions convened by the American Psychological Association that examined both the benefits and challenges of AI, with particular attention to impacts on young users. The promise of AI for health care, generally, remains undisputed among experts, and the application of those benefits for mental health therapies (MHT) is being explored. These are important issues for rural spaces, which face mental health care provider shortages and in which the choice of fewer providers can compound the social stigma often perceived by those seeking help. AI-driven mental health therapy applications can serve as valuable tools in the therapeutic arsenal.
AI-powered conversational agents can serve as a gateway or frontline interaction to human therapeutic intervention. For prospective patients, they offer easy access, unencumbered by otherwise discouraging financial or social barriers; for therapists, the tools can digest and triage conversations and enable providers to broaden their patient portfolio. NTCA addressed these issues as well targeted use cases for younger populations in our NTCA AI white paper (see page 6 there). And while I personally (and professionally) champion the use of technology for meaningful applications (see our Smart Rural Community papers on XR modalities for rural telehealth, or technology to protect survivors of domestic abuse), it is also necessary to assess the potential costs alongside the apparent benefits.
The deployment of AI for MHT implicates known hazards that merit careful consideration. These include the natural tendency of AI models to affirm user input, which increases the risk of confirmation bias if the tool leans toward affirming behaviors rather than, where appropriate, coaching users toward change. And when it comes to adolescent users, different risk factors demand attention.
Teenagers are increasingly turning to AI companions precisely because those inanimate apps are non-judgmental and readily accessible. Popular "general purpose" large language models (LLMs) like ChatGPT make no explicit claims about addressing wellness or mental well-being, yet they are being used precisely for that purpose. Even platforms like Character.ai, which are built explicitly to function as an "AI friend," often lack the input of subject matter experts in behavioral health. All of this comes at a time when Pew Research data demonstrates declining teen usage of traditional social platforms like Facebook and WhatsApp – spaces in which users actually interact with other humans (the merits and hazards of social media can be referred to a different conversation). In contrast, two-thirds of teens are now using chatbots, and Common Sense Media reports that nearly 75% of teens engage with AI companions. Perhaps more compelling, however, are survey results revealing that approximately 33% of teens use chatbots for emotional support and friendship, and that an equal proportion of teens have turned to chatbots instead of humans for help.
Coupled with threshold concerns that AI companions are designed to be affirming by default, this level of engagement raises significant questions. The inclination of chatbots to affirm user positions does not support cultivation of foundational relationship building experiences in developing minds (in fact, experts advise users to actively challenge predictions made by LLMs even when engaging for otherwise basic research or other tasks). Michael Robb, head of research at Common Sense Media, explained that AI companions do not present users with situational conflict. Accordingly, without training in managing conflict (similar to "getting the reps in" during practice), young people will not develop the skills necessary to communicate and resolve differences effectively with others "on the field," whether that field be school, family, or the workplace.
Robb’s observation connects to something we have discussed previously, namely, the critical role of soft skills in an increasingly tech-focused economy. For example, soft skills constitute a major value proposition for esports, which demand not only game-focused proficiency but also the ability to communicate and collaborate effectively. But if soft skills are what will differentiate skilled workers in the future labor market, what risks might develop as growing numbers of adolescents rely increasingly on AI companions during their formative developmental years? This is no longer simply losing the art of conversation; this risks shortfalls in the ability to communicate effectively, including in instances of disagreement.
And hence “The Sorcerer's Apprentice.” We are all familiar with Goethe's narrative, retold in Disney's "Fantasia": The sorcerer leaves his apprentice to complete mundane chores, including fetching water. Tiring of the task, the apprentice conjures a broom to perform the work autonomously. But the broom fetches too much water. And when the apprentice breaks the broom in an attempt to halt it, the broken pieces each become a new separate broom, continuing to fetch water and ultimately causing a catastrophic flood.
Is it a perfect analogy? Perhaps not, but it serves as a cautionary tale about the unintended consequences of powerful tools deployed without adequate supervision or controls. Technology possesses demonstrable power to help youth achieve well-being and developmental success. But technology can innovate faster than our capacity to grasp its implications, and the lessons learned from the effects of social media on adolescent mental health counsel careful attention as AI usage among teens continues to increase.
As locally operated ISPs help schools and MHT providers deploy the right tools to address local needs, rural broadband providers can help their other local leaders avoid proverbial floods of unintended consequences - understanding that AI, while valuable, is an apprentice best supervised.
While challenging, the task at hand can align directly with opportunities that are promoted in NTCA's Smart Rural Community program- specifically, the recognition that locally operated ISPs are uniquely positioned to help community leaders implement technology solutions that work best for their communities. In this case, rural ISPs with boots-on-the-ground understanding of local healthcare and MHT conditions (including access and financial and/or social barriers) can help schools and MHT providers avoid proverbial floods of unintended consequences - understanding that AI, while valuable, is an apprentice best supervised.