DATASCI 101: Introduction to AI Applications

Lecture 22: AI and Wellbeing: The Attention Economy, Mental Health, and the Environment

Danilo Freire
danilo.freire@emory.edu

Department of Data and Decision Sciences
Emory University

Welcome back!

Recap of last class

  • AI targets cognitive tasks, not just physical ones
  • Automation displaces tasks, not whole jobs (Acemoglu & Restrepo)
  • Careful estimates find no effect on earnings or hours so far
  • Young workers in AI-exposed jobs take the first hit
  • Today: how AI affects attention, mental health and the environment

Source: China Daily

Lecture overview

Today’s agenda

Part 1: The attention economy

  • “Your attention is the product”
  • How recommendation algorithms work
  • Filter bubbles and teen screen time
  • What platforms are changing

Part 2: AI and mental health

  • Who gets mental health care?
  • Chatbots in therapy and the evidence
  • Risks and hard limits
  • Augmentation, not replacement

Part 3: AI and the environment

  • Energy and water costs
  • Footprint in context
  • AI as a climate tool

We will pause for discussion after each part

There are no right answers to the questions I put on screen. I’m curious what you think!

(Sad) meme of the day

The attention economy

What is the attention economy?

  • Herbert Simon (1971): information-rich environments create attention-poor ones
  • Your time is the scarce resource (Becker, 1965)
  • Platforms sell your attention to advertisers, as TV and newspapers did
  • Tristan Harris (ex-Google): “the race to the bottom of the brainstem”
  • Algorithms maximise engagement, not truth or wellbeing
  • The Attention Merchants (Wu, 2016) traces this from 19th-century papers to social media

Source: The Atlantic

Does the incentive structure produce good outcomes regardless of intent?

How recommendation algorithms work

  1. You watch, like, or share something
  2. It builds an embedding of “you” from your behaviour (lecture 06)
  3. It compares your vector to what similar users watched next
  4. It recommends what they watched
  5. You watch, the embedding updates, and the loop repeats
  • Engagement = watch time, clicks, shares
  • TikTok learned bots’ interests in under 2 hours from how long they lingered (WSJ, 2021)
  • YouTube autoplay chains videos by design

Source: Data Science Dojo

Every swipe teaches the algorithm something about you

These systems have no concept of “healthy” consumption. They optimise the objective they are given

Filter bubbles: real or exaggerated?

  • Eli Pariser coined “filter bubble” in 2010: feeds that only confirm your views
  • Two large experiments suggest it is probably overstated
  • Nyhan et al. (2023, Nature): cutting like-minded Facebook content barely changed polarisation
  • Guess et al. (2023, Science): a chronological feed made little difference to attitudes
  • Effects beyond polarisation (anxiety, body image, sleep) are still debated

Good news on polarisation. Mental health is a separate, contested question

Engagement vs wellbeing

  • YouTube moved from clicks (2011) to watch time (2012), then added satisfaction surveys
  • Outrage spreads: each moral-emotional word raised retweets ~20% (Brady et al., 2017, PNAS)
  • Algorithms learn this and show more of it
  • Facebook’s own research (2021, WSJ): Instagram made body image worse for teenage girls
  • Facebook kept the feature. Fixing it would have cost engagement
  • Revenue and user wellbeing point in opposite directions

Companies answer to shareholders. When engagement and wellbeing conflict, engagement wins

Teens and screens: what the data says

How large is the effect?

Factor Association with wellbeing
Screen time ~0.4% of variance
Wearing glasses ~1.5x screen time (one dataset)
Getting enough sleep 1.7x to 44x screen time
Eating breakfast 2.4x to 31x screen time

Small effects still matter at population scale, but other factors matter more

What platforms are changing

Platform Change Scope
Instagram “Snooze suggested posts” Opt-in
TikTok 60-min daily limit for under-18 Bypassable
YouTube “Take a break” / “Bedtime” reminders Default-on for 13-17
Meta Teen Accounts with safer defaults Under-18
  • Most are opt-in or reach a minority of users. The ad model is unchanged
  • They also help companies look responsible to regulators
  • US Surgeon General: youth mental health advisory (2023), then a call for warning labels (2024)

Source: The Verge

Probably both concern and PR. We’ll know if these changes survive once regulators look away

China’s experiment: opting out of the algorithm

What the regulation requires:

  • Users must be told they are being profiled
  • An opt-out from personalised recommendations
  • Algorithms must not exploit addictive behaviours
  • Platforms must label AI-generated content
  • Special protections for minors

A real test of what happens when users can choose their feed. Results are still coming in

Discussion: designing for wellbeing

If you were designing a recommendation algorithm and your bonus depended on user wellbeing instead of watch time, what would you change?

  • What metric would you optimise for? How would you measure “wellbeing”?
  • Would your platform still be profitable?
  • Would users stay, or move to a competitor with more dopamine hits?

AI and mental healthcare

Who gets mental health care?

  • WHO: 75% of people with mental illness in low- and middle-income countries get no treatment
  • ~0.1 psychiatrists per 100,000 in low-income countries
  • Even in wealthy countries:
    • US wait for a first psychiatric visit: median 67 days (five states), longer if rural
    • A session costs $120-230 on average, depending on the state
    • Stigma stops many from seeking the care that exists
  • When the alternative is no care at all, AI tools look different

Suicide-prevention programmes by country income (WHO)

Source: WHO Mental Health Atlas (2024)

“Better than nothing” is a low bar, but for millions it is the only option

AI as a therapeutic tool

What the research shows

  • Opel and Breakspear (Science, 2026): AI “may reduce care inequities when deployed responsibly”
  • Both “may” and “responsibly” are big caveats
  • Small RCTs show short-term relief for mild-to-moderate anxiety and depression
  • Very few studies run past 8-12 weeks, and dropout is high
  • Publication bias: positive results get published, negative ones mostly don’t
  • The evidence is much thinner than the marketing

Evidence by level:

Level Status
Long-term RCTs Almost none
Short-term RCTs Mixed, small samples
Observational Positive signals
User self-reports Generally positive
Marketing claims Very positive

Risks

Documented risks:

  • Users grieved like a lost relationship after Replika’s 2023 feature change
  • Chatbot data is rarely protected like clinical notes
  • LLMs showed stigma and mishandled crises (Moore et al., 2025)
  • A chatbot cannot call an ambulance
  • Most are apps, not medical devices. A few states (Illinois) ban AI therapy

Compared with human care:

  • Human therapy isn’t fully private either: notes, supervision, insurance coding
  • Weigh harmful responses against no care at all, the norm in poorer countries
  • AI may give governments and employers an excuse to avoid funding real care

The status quo also harms people

What LLMs can and cannot do

LLMs cannot:

  • Keep reliable long-term memory of your care
  • Verify what you tell them, as a therapist does over months
  • Provide legally enforceable confidentiality
  • Diagnose or prescribe medication
  • Read body language, tone or facial expression

LLMs can:

  • Be available at 3am
  • Be patient, never judge, never tire
  • Explain anxiety, depression and coping strategies
  • Scale to millions at near-zero marginal cost

Chatbots and therapists have different strengths

Questions to think about

It’s 3am, you can’t sleep, you feel anxious. Would you talk to a chatbot? What would it need to do for you to trust it?

If AI therapy lets a government say “we’ve addressed mental health” without funding real services, is that a net positive or negative?

Augmentation, not replacement

Credible use cases:

  • First contact: triage with less stigma, then referral to professionals
  • Between-session support: coping strategies and mood tracking (the Habicht et al. model)
  • Stepped care: AI for mild symptoms, humans for moderate-to-severe
  • Admin: notes, scheduling and finding services, to free therapist time
  • OpenAI added crisis signposting to ChatGPT in 2025: people already use it this way
  • Who decides what “responsible” looks like?

Stepped care in practice:

Mild symptoms
  → AI triage + self-help tools

Moderate symptoms
  → AI support +
    case management referral

Severe symptoms
  → Human professional care,
    AI for admin support only

The technology moves faster than the evidence, the regulation and the training

AI and the environment

The energy cost of training AI

  • Training uses a lot of energy, and companies rarely disclose it
  • Patterson et al. (2021): training GPT-3 produced ~552 tonnes of CO₂:
    • 368 return flights London-New York
    • 120 cars driven for a year
    • The annual footprint of ~35 Americans
  • GPT-4’s footprint is undisclosed, and outside estimates vary widely
  • Inference (billions of queries) probably dwarfs training
  • Microsoft’s emissions rose 29% between 2020 and 2023, partly from AI
  • Google (2026): emissions up 81% since 2019, driven by AI data centres

Training is a one-off. Inference is the number that grows with adoption

AI and water

  • Li et al. (2023): training GPT-3 used ~700,000 litres of cooling water
  • GPT-3: ~500ml per 10-50 responses, while Google reports 0.26ml per Gemini prompt
  • 70% of Americans would oppose a local data centre (Gallup, 2026)
  • US data centres use 627m gallons a day, less than US golf courses
  • About 75% of a data centre’s water footprint comes from generating its electricity, not on-site cooling
  • Clusters in Phoenix, Las Vegas and Santiago, Chile sit in water-stressed regions

Source: Axios (2026)

Aggregate figures hide local impacts, such as how much water a data centre takes from its region

AI’s footprint in context

Current estimates (IEA, 2025):

Sector Share
Aviation ~2.5% of global CO₂
All data centres ~1–1.5% of electricity
AI specifically ~0.2–0.3% of electricity
  • AI’s footprint is smaller than aviation’s or steel’s
  • AI servers’ power use may grow ~30% a year (data centres: ~15%)
  • IEA: data centres could double to ~945 TWh by 2030, about Japan’s total
  • Jevons paradox: efficiency raises use, cancelling the gains
  • Gemini’s energy per prompt fell 33x in a year, but queries multiplied

Adding between one Sweden and one Germany by 2026

Source: IEA (2024)

AI’s share is small today but growing fast

AI as a climate tool

AI may also cut emissions:

The net impact is unclear

Discussion: the environmental trade-off

You run an AI startup. A journalist asks about your carbon footprint. What do you disclose, and what do you leave out? Why?

  • Do you report training costs, inference costs, or both?
  • Do you compare your footprint to other industries (aviation, streaming)? Is that honest or deflecting?
  • Would full transparency help or hurt your business?

Summary

Main takeaways

  • Attention economy: feeds optimise engagement. Filter bubbles look overstated, mental health effects are debated

  • Mental health: a huge treatment gap. Short-term promise, little long-term evidence

  • Environment: inference matters more than training. AI may cut emissions elsewhere, net impact unclear

  • Across all three: the business model rarely rewards wellbeing

  • No tidy answers: you will face these trade-offs throughout your career

…and that’s all for today! 🎉