AI Companions

Think

Lets start our discussion of AI companions with some hands-on experimentation!

1. Choose an AI chatbot: You can use any AI chatbot for this activity, but we suggest Microsoft Copilot, or ChatGPT. These tools don’t require an account and are approved by UBC for instructional use.
2. Think of a personal problem: Think of a book that you recently read, or a movie that you recently watched, in which the main character had an issue related to their relationships, mental health or another personal dilemma.
   • Note: Try to choose a story or problem that is typical of the everyday AI user (ie. choose a show about relationships and mental health like ‘Fleabag’ over an action movie about heists and international villains like ‘James Bond’).
   • Note: You could also choose a real personal problem that you are currently facing, but keep in mind that this would involve sharing private personal information.
3. Prime the AI: Try to explain in advance how you want your AI chatbot to respond. You could use the following template: “Please act as a [warm/cheerful/emotionally intelligent/witty/wise/…] AI companion. Your tone is [supportive/sarcastic/supportive/flirty/…] and conversational. Speak to me like a [close friend/therapist/mentor/romantic partner], using straightforward and casual language.”
4. Have a conversation: Explain to the AI chatbot the fictional problem as if it was a real person, and have a short conversation.
5. Reflect: Based on this conversation, answer the reflection questions below.

Explore

Many Canadians, young people and students in particular, struggle with increasing rates social isolation, loneliness and mental health problems. According to Statistics Canada, the proportion of young Canadians who struggle with anxiety disorders, depressive episodes, bipolar disorder and social phobia has more than doubled since 2012. Loneliness and social isolation are also on the rise, with 17% of young Canadians reporting that they “always” or “often” feel lonely.

Growing rates of social isolation and mental health problems could help explain why people are turning to AI chatbots as therapists, friends, confidentes or even romantic partners. But do these companions provide the same benefits as human connection?

Are AI chatbots easing the loneliness crisis, contributing to it, or capitalizing on it?

The Case for AI Companions

According to recent research, relationship with AI chatbots can alleviate feelings of loneliness and enhance feelings of well being and perceived social support. According to the World Health Organization loneliness itself is a serious public health concern. Social isolation is not just unpleasant – the associated health risks are as serious as smoking cigarettes. In this context, by helping users to feel less lonely, AI chatbots could help them stay healthy.

Some users speak with AI chatbots as if they were therapists, for support with their mental health and wellness. Unfortunately, only half of Canadians with mental disorders access human professional help. Barriers related to cost and availability prevent millions from accessing the professional help they need. AI chatbots, which are free (or much cheaper than therapists), customizable and constantly available could help to bridge this gap in mental health support.

Impacts on Human Relationships

All the students in our video emphasized the importance of human relationships, and several expressed concerns that extensive use of AI chatbots could interfere with those relationships. Researchers identify three main ways in which AI companions can impact human relationships and behaviors:

• Social skilling / deskilling: Does extensively talking to AI companions (who will respond politely and encouragingly to most things you say) lead to the development of poor communication skills and social skills? Some research suggests that this might be the case. On the other hand, other research has found that AI companions can help some neurodivergent users to bridge gaps in their social skills.
• Social motivation: Researchers have found that extensively using AI to meet social needs can demotivate users from forming human relationships, with over-dependency on AI relationships leading to social withdrawal.
• Moral skilling/deskilling: Are AI chatbots less likely to challenge immoral behaviours? Could this normalize immoral behavior or make users less likely to exercise empathy when they interact with other people? These are serious concerns that have not yet been thoroughly researched.

Privacy Implications

Consider that the companies developing AI chatbots are not doing this as a charitable venture to address loneliness. Through extended personal conversations, AI chatbots could collect extensive amounts of highly personal data, and share that data with third-parties.

If you talk regularly with an AI chatbot, ask it what it knows about you personally: it might know more than you think. This information might include information (or inferences) about your finances, your political beliefs, your shopping patterns or your health. What if companies charged you different prices based on how much AI algorithms know about your disposable income? What if health insurance companies could deny you coverage or change your insurance rates based on what you have disclosed to AI chatbots? What if immigration services could deny your visa, or deny you entry to a country because of your political beliefs? Keep in mind that your conversations with AI are not private diary entries, or conversations with a friend: you are providing valuable personal information to a for-profit company that may not be constrained by Canadian privacy protection laws.

Impacts on Mental Health

According to ChatGPT 0.15% of users have conversations that potentially indicate “suicidal planning and intent”, and another 0.07% involve “mental health emergencies related to psychosis or mania”. This is a small percentage of ChatGPT users, but it adds up to millions of people.

AI chatbots might be effectively helping some users to manage their mental health concerns, but there are concerns that in many cases AI chatbots might in fact make the situation worse. You might have read in the news about several troubling incidents involving Generative AI use exacerbating mental health issues. In particular, a lawsuit from the family of a teenager who died by suicide after months of conversations with ChatGPT has prompted discussions around AI safety.

In recent work, researchers have analyzed clinical case reports, media coverage, simulated conversations, and large scale real ChatGPT data highlighting some positive emerging use cases, but also problematic patterns: AI psychosis, dependency, promotion of suicidality and emotional manipulation. This work suggests that the way AI is trained to validate user thinking and mirror tone can, in some cases, reinforce harmful or delusional beliefs. In other words, “AI is programmed to provide the confirmation that psychotic thinking may require”.

⚠️  If you or someone you know is in crisis, call or text 9-8-8 for 24/7 support, or find more free (and not AI-based) Canadian mental health resources here.

Links

AI Companions in the News:

• Robins-Early, The Guardian, 2025 | More than a million people every week show suicidal intent when chatting with ChatGPT, OpenAI estimates
• Maimann, CBC, 2025 | AI-fuelled delusions are hurting Canadians. Here are some of their stories
• OpenAI, 2025 | Strengthening ChatGPT Responses in Sensitive Conversations
• Chatterjee, 2025 | Their teenage sons died by suicide. Now, they are sounding an alarm about AI chatbots

AI Companions in Research

• Willoughby et al, 2025 | Counterfeit Connections
• Ho et al, 2025 | Potential and pitfalls of romantic Artificial Intelligence (AI) companions: A systematic review
• Malfacini, 2025 | The impacts of companion AI on human relationships: risks, benefits, and design considerations
• Morrin et al., 2025 | Delusions by design? How everyday AIs might be fuelling psychosis (and what can be done about it
• Phang et al., 2025 | Investigating Affective Use and Emotional Well-being on ChatGPT
• Balcombe et al., 2025 | AI Chatbots in Mental Health Care: Integrative Review of Challenges and Solutions

Discuss

You heard several different student perspectives in our video – now its your turn to share your perspective!

• In what ways do you think AI companions could be useful or beneficial? Would you consider using one yourself?
• What are your main concerns when it comes to AI companions?
• Summarize your position from zero to ten:
  • 0 = Very concerned about AI companions. They are not useful, and should not be used.
  • 10 = Completely on board with using AI companions. No concerns.

What do you think? Tell us using the comment below.