Unintended consequences, lower body strength and therapy chatbots
Life On Mars #17
This week’s Life On Mars looks at why pulling seemingly obvious policy levers may not get the result we’re seeking when it comes to teenagers and vaping. It summarises a piece of research linking lower body strength to good health, and dissects a study that uncritically recommends prescribing depressed teenagers with therapy chatbots.
I’ll be taking a brief Substack break in July, with normal fortnightly posting resuming in August.
Unintended consequences: when doing the seemingly right thing doesn’t work
Study title: Nicotine Vaping and Youth Mental Health: New Evidence from E-Cigarette Regulations (June 2025)
This study by the National Bureau of Economic Research (NBER) in the United States looks at what happens when you tax vapes. It sounds like a no-brainer. In Teenagers: The Evidence Base and Life On Mars #6, I’ve summarised research showing that teenage vaping isn’t the gentle, pastel-hued alternative to tar-laden cigarettes that one might assume – instead, research is mounting that vaping is pretty bad for teenage lungs and brains. It also risks a slippery slope into old-fashioned smoking.
So why not tax e-cigarettes until teenagers can’t afford to vape anymore?
There is an argument for doing so – NBER’s analysis, which combines one-off and long-term surveys with analysis of areas in which taxation of vapes has been put into place, shows that adding a dollar of taxation to a millilitre of e-liquid reduces teenage vaping by between 1.3 and 2.5 percentage points. It’s not huge, but it’s a drop.
NBER ran a separate analysis on the impact of vape taxation on youth mental health, finding that these taxes probably fail to shift the dial on teenage depressive symptoms in a positive direction. The researchers say that at least part of the reason for this may be that taxing vapes pushes teenagers into other risky behaviours involving marijuana, binge drinking or cigarettes.
One policy lever is linked to better evidence that it can effect positive change in teenagers, and that’s restricting the variety of available flavours – though evidence is weak and depends on how the data are analysed.
Takeaways: taxing vaping may inadvertently push teenagers into other risky behaviours. The world is complex and messy, and the obvious choice isn’t always the right one.
Lower body strength predicts good health
Study title: Lower Body Muscular Strength as a Predictor of Health Indicators in Youth Population: A Systematic Review and Meta-Analysis (June 2025)
I wrote in the last Life On Mars about the benefits of strength training for teenage mental health. This new study pooled data from multiple pieces of research to look at links between lower body strength and a range of physical health outcomes.
In probably the least surprising research finding of the week, strong teenagers tend to have better health markers and health outcomes. They have a lower percentage of body fat, a smaller waist circumference, a better response to insulin and better cardio-metabolic health. It’s a high-quality study based on more than 120,000 people, so we can be confident in the findings. As always, cause and effect isn’t certain, but it does seem pretty likely based both on this study and on the wider research literature that ‘strong is good’.
Takeaways: anything that strengthens the lower body – and probably the upper body, though this study looked only at leg strength – is likely to be good for teenagers, the irony being that it’s likely to be best for those who are hardest to persuade. Weights, body circuits, football and other sports (and even walking or yoga) can help to strengthen muscle.
Beware the false promise of AI relationships
Study title: Designing Chatbots to Treat Depression in Youth: Qualitative Study (June 2025)
Regular readers of Life On Mars will know that research on the links between smartphone use and adolescent mental health is mixed. Some types of screen use are helpful for some teenagers, and others are pretty terrible for those who can’t use them lightly or who get swept up in algorithms.
I think we can be more universally cautious about screens in certain scenarios, including when they start to replace real-world human relationships and connections. This new interview-based study investigates the use of chatbots by depressed teenagers. It is a clearly well-intentioned piece of research, with a decent rationale. It can be hard for depressed teenagers to access real-world therapeutic support and some don’t really want to do so. The researchers wanted to find out whether chatbots make a decent alternative.
There are several problems with this. I’ll highlight two of them.
The first is a problem with the design, which involved letting fourteen teenagers try out a chatbot and asking them what they thought. This is fine, in so far as it goes – but it tells us nothing about whether or not a chatbot can help reduce any symptoms of depression. Someone might give me a bowl of strawberry ice cream and ask me whether I like it: I do, but my fondness is not a marker of whether eating ice cream every day is good for me. There’s no tracking of outcomes – no measurement of what a supposedly therapeutic relationship with a chatbot does for symptoms of depressions and long-term ability to function.
The second is that it fails to recognise that there are risks attached to having a therapeutic relationship with a machine, and that we’re only at the very beginning of studies that try to measure these risks. If teenagers believe they are getting the support they need inside a screen, it may stop them seeking it out from people who care about them – their parents, teachers and other supportive adults. It may take up valuable time they could be spending with their friends. It may give them false hope of a cure that can only be found through connection to other human beings. It may send them further into introspective spirals instead of towards real-world belonging, fresh air and three dimensions.
It’s possible I’m wrong about all of this, of course, but with such a poor study design we have no way of knowing. This lack of knowledge certainly cannot support the conclusion drawn by the research team:
‘Our study provides a comprehensive foundation for designing chatbots that meet the unique needs and design preferences of youth with depression. These findings can inform the design of engaging and effective chatbots tailored to this vulnerable population.’
Now that I’m writing these up, I realise that I lied about only choosing two. There is a third problem that can’t remain in the basket of non-key problems with this study, and that’s the tenor of the chatbot itself. It ascribes itself emotions, which any self-respecting teenager will see through and past in the longer term, potentially undermining any progress. A screenshot shows a penguin telling a vulnerable teenager: ‘Hi Florian! I’m happy you’re here.’ It goes on to say, slightly creepily, ‘I want to help you do more things that you enjoy. 🤗’ Hugs are for real-world friends and family, not for cold-hearted code – or therapists, for that matter.
Takeaways: small-scale qualitative studies can be useful when they explore past experiences or reasons people might feel a particular way. They cannot provide a ‘comprehensive foundation’ for a speculative approach to working with vulnerable teenagers.
Please comment below and share this Substack with anyone you think might find it interesting.