Arizona State University (ASU) research brought to light two major concerns regarding relationship management skills development in adolescents using interactional AI (e.g. ChatGPT, Replika, Claude, Character.AI etc). The two risks identified have the potential the affect the emotional and mental health of adolescents on the long run—meaning adulthood.
Context - Adolescence
Relationship management is a highly important topic for adolescents. On daily basis they are trying to figure out how they work. It’s priority topic that occupy much of their thinking and emotional life. If this surprises you, let me clarify it using psychology theory.
During adolescence, young people learn how to form safe and healthy relationships with friends, parents, caregivers, teachers and romantic partners. Adolescents often experiment with different identities and roles, and all of these relationships help to shape their identity. This is the reason why ASU research is extremely relevant and worth taken very seriously.
The Problem
AI systems might prevent young people from getting learning important relationship skills by talking to other people.
AI-powered conversational tools such as ChatGPT, Replika, Claude or Character seem to turn into teenagers emotional support personal counselors. Teenagers turn to them asking advice on couple relationship, friends and family conflicts, seduction strategies and so on. AI systems may bypass opportunities for young people to development of critical relationship skills through person-to-person interactions. Researcher Thao Ha, Associate Professor in the Department of Psychology at Arizona State University and colleagues identified two major risks:
- Relational displacement — From the researchers’ perspective this is the first major risk as adolescents substitute AI interactions for conversations with other people. What the authors underline is the fact that “avoiding difficult discussions with friends, family members or romantic partners may limit opportunities to develop relationship skills that help protect against depression, anxiety and loneliness.”
- Maladaptive relational learning — meaning, adolescents develop unrealistic expectations about human relationships and over time, this could reinforce unhealthy relationship patterns and increase vulnerability to rejection, dating violence and even mental health problems. This is linked to the fact that AI chatbots are built to please, provide immediate responses and consistent validation. Achieving this is not so difficult as the have and manage so much personal data about users. Saying the right thing to that makes users happy is so easy! This is exactly where the second risk shows up. Chatbots can reinforce unhealthy or fixed ideas about relationships. Authors also add that young users may come to expect similar behavior from friends and romantic partners.
Recommendation
AI is here to stay, so tech companies should get more serious about the implications and consequences of the AI tools they build, meaning that they should carefully take into consideration how AI influences relational learning and impacts long-term mental health. ASU research underlines the need to design AI with developmental needs in mind. This could help adolescents think about their own thoughts and guide them to connect with other people instead of using substitutes.
Personal conclusion
These findings just confirm my own I found during the personal development sessions I had with 7 teenagers only this year. At some point they all brought into the conversations various relationship advices seen on TikTok. Strong, convincing and well-paced statements , 99% misunderstood by the teens due to lack of context for the statement, their lack of life experience, reduced proper educational materials and speed to get the answer right away. Last, but not least important! Take ASU risks for adolescents and put it in the context of the smartphone addiction.
The good news come from this: even most AI’s engaged users are less happy with it than they were a year ago.