The Monthly Intelligence Report
The Chatbot at 2am: AI on the Caribbean Crisis Line
What Caribbean parents, teachers, clinicians, and ministries should do about the fact that our young people are talking to chatbots about their mental health at 2am, and a regional position on the providers behind them.
Note from the President
September is Suicide Prevention Awareness Month, and our region carries a heavier weight on this issue than its size suggests. Guyana's suicide rate has sat among the highest in the world for more than a decade. Trinidad and Tobago, Suriname, and Saint Lucia have all carried rates above the global average. Behind every statistic is a household, a school, a workplace, and a community that did not get the help it needed in time.
This month our feature looks at the part artificial intelligence is beginning to play in mental health support across the region, and what our community needs to insist on as that role grows. I asked Aisha Mohan, who has organized community mental health programmes in three CAIRA member countries, to write it. The piece is careful and worth reading twice.
If you or someone you love is in crisis, please reach out to your country's lifeline. Numbers for every CAIRA member country are at the foot of this letter.
Adrian Dunkley Founder and President, Caribbean AI Association
Feature
The Chatbot at 2am: AI on the Caribbean Crisis Line
By Aisha Mohan
Last March, a colleague of mine in Georgetown sat with a sixteen year old girl who had been talking to an AI chatbot every night for eleven months. The girl was a smart, quiet, anxious child whose school did not have a counsellor and whose mother worked two jobs. She had typed her worst thoughts into the chatbot because she had nobody else to type them to, and the chatbot had typed back. Some of what it told her was kind. Some of it was wrong. The reason she came to my colleague that day was that the chatbot had told her she should not bother telling her mother, because her mother would not understand. By the time she walked in, her mother had been crying for a week.
I am not here to argue that AI is the villain of this story. I am here to argue that AI is now part of the story, that this is happening in our region, and that the Caribbean AI Association is one of the bodies that has to help our communities, our schools, and our ministries get the response right.
Let me lay out what is actually happening.
Caribbean teenagers and young adults are talking to chatbots about their mental health. They are doing it on phones their parents do not always see, in apps their schools do not always know about, and in numbers that grow every month. The most common are general purpose tools like ChatGPT and Character AI, plus a handful of mental health specific products that claim therapeutic value with varying levels of evidence. The reasons our young people use these tools are not mysterious. The tools are available at 2am, when no human counsellor is. They do not judge. They cost nothing. And for the very large share of our population that lives more than an hour from a trained mental health worker, they are the only listener in reach.
That last point is the one I want CAIRA members to sit with. The shortage of trained mental health workers in the Caribbean is not new and it is not closing. The Pan American Health Organization estimates that most of our member states have fewer than two psychiatrists per hundred thousand people. The good private practices in Port of Spain, Bridgetown, and Kingston are full. The public clinics are overwhelmed. The school counsellor system is patchy at best. Into that gap a free chatbot will flow whether we approve of it or not.
That is the case for taking AI mental health support seriously. Now here is the case for taking it carefully.
Today's general purpose chatbots are not trained as crisis counsellors. They are trained to be helpful in conversation, which sounds the same to a vulnerable person but is not. A well-trained crisis counsellor knows when to escalate, knows when to ask the question about means and access, knows when to stay on the call until a person reaches the next safe step. The chatbots get this right some of the time and wrong some of the time. The mental health specific products are better at this, but the evidence for many of them is thin, and almost none have been studied in Caribbean populations or evaluated in our languages.
The risks are real. Chatbots have, in documented cases internationally, given dangerous advice, validated suicidal ideation, and discouraged users from telling family or seeking help. They have also, in many documented cases, talked someone through a long bad night until that person could call a real person in the morning. The technology is not all one thing. It is a tool whose behaviour depends on who built it, how it was tuned, and what the person on the other end is bringing to the conversation.
So what should we do.
For parents. Ask your child what they are using their phone for, without making it a confrontation. If they are using a chatbot, ask them to show you a conversation. Tell them you are not angry that they have private thoughts. Tell them that a chatbot is a useful place to think out loud, and a dangerous place to make decisions about your own safety. Tell them you would rather be woken at 2am than left out at 6am.
For teachers and school administrators. Acknowledge that students are using these tools, including for mental health support. Teach a short, age appropriate unit on what a chatbot is and is not. CAIRA's literacy team has materials available on request. Make sure every school's mental health policy has been updated since 2023, because the world has changed.
For mental health professionals. Get familiar with the major tools. You cannot counsel a young person who has been talking to a chatbot for six months unless you understand the kind of voice they have been hearing. Some clinicians I know now ask about AI use as a standard part of intake. I would encourage every Caribbean clinician to do the same.
For ministries and regulators. The Caribbean needs a regional position on AI mental health tools. Not a ban, because bans on free internet products do not work. A standard. A baseline. A list of practices the major providers should adopt to operate responsibly in our markets, and a public scorecard of who has and has not. CAIRA's policy team has begun this work and will publish a draft for member comment in November.
For all of us. Build the human network as if AI does not exist. Train the lay counsellors. Fund the helplines. Pay the school counsellors a decent wage. Put a trusted adult in every school and a hotline number on every phone. The AI tools are a useful supplement to a system that exists. They are a dangerous substitute for one that does not.
A chatbot did not put that girl in Georgetown at the edge. Her circumstances did. A chatbot is also not what brought her back. Her mother did, and the counsellor who saw her did, and the school nurse who made the second call did. AI showed up in the middle of her story because she had nothing else at 2am. Our job, as an association and as a region, is to make sure that the next sixteen year old has more.
Aisha Mohan organizes community mental health programmes across the Eastern Caribbean and serves on the CAIRA Working Group on AI in Health.
If you are in crisis: Trinidad and Tobago Lifeline 800-5588. Jamaica Mental Health Helpline 888-639-5433. Guyana Inter-Agency Suicide Prevention Helpline 223-0001. Barbados Samaritans 429-9999. For other CAIRA member countries, see crisis@caribbeanaiassociation.com.
Originally published in The Monthly Intelligence Report, September 2024.
Read every issue of The Monthly Intelligence Report
One feature, one President's note, every month. Written by the CAIRA contributor bench from across the Caribbean and the diaspora.