BusinessMarkets, money & strategy
UCL's Henrietta Moore argues AI is adding a 'cognitive tax' on low-paid workers and calls for universal basic services
UCL's Henrietta Moore argues that AI harms low-paid workers mentally before it takes their jobs. Unstable pay, housing and debt drain the attention they need to retrain, and AI scheduling and tool supervision add to that load. She proposes universal basic services.
From the The Guardian · AI article of November 20, 2026 and cited sources: Moore uses 'cognitive tax' as her own name for the mental load of financial insecurity, not as a new empirical finding. [1]
SOURCES[1] AI is imposing a hidden ‘cognitive tax’ on the lowest-paid… theguardian.com
AI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Moore uses 'cognitive tax' as her own name for the mental load of financial insecurity, not as a new empirical finding. [1]
- Original source
- The Guardian · AI article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 20, 2026 article published
- Covered
- Nov 20 edition
- Evidence
- 17 sources across 17 sites
- 1AI is imposing a hidden ‘cognitive tax’ on the lowest-paid…theguardian.com
- 2NBER: Do Financial Concerns Make Workers Less Productive?nber.org
- 3Route Fifty: Guaranteed income study resultsroute-fifty.com
- 4WEF Future of Jobs 2025weforum.org
- 5Walmart AI tools announcementcorporate.walmart.com
- 6BCG: AI brain frybcg.com
Reach: not measured yet

Moore uses 'cognitive tax' as her own name for the mental load of financial insecurity, not as a new empirical finding. She says AI raises this load through less predictable schedules and the strain of supervising AI tools. Her remedy is universal basic services such as free childcare, housing, transport and digital access. She also wants the UK to promote cognitive-load tests during its G20 presidency. [1]
The research linking scarcity to weaker thinking is strong. A meta-analysis covering more than 111,000 people found a moderate negative effect on cognitive performance, though education accounts for much of it. In a field experiment, Indian workers paid earlier were about 7% more productive because they made fewer attention errors, and gains topped 10% for the poorest workers. [9] [11] [2] [12]
Evidence for the AI-specific part is indirect. In a BCG survey, workers with heavy AI-oversight duties reported 14% more mental effort and 12% more fatigue, and perceived productivity fell when people used more than about four tools. Claims that AI scheduling cuts hours come from individual worker accounts, while Walmart presents its tools as giving staff more flexibility. [6] [13] [17] [14] [5]
For employers, the implication is to treat schedule predictability and per-person oversight load as design measures alongside labour cost. A large guaranteed-income study found that stress relief faded after the first year, which suggests cash alone may not be enough. However, no reviewed trial shows that services work better than cash. [3] [4] [17]
Causal studies show that financial strain reduces attention and productivity. No study yet measures AI's specific effect on low-paid workers.
Read the full assessment
If the link holds, scheduling and agent rollouts designed only to cut costs could undermine the reskilling that employers expect most workers to need by 2030.
Executive brief
The argument is that AI's first harm to low-paid workers is a mental load that comes before any job loss. In a Guardian comment piece, UCL's Henrietta Moore calls this load a "cognitive tax": unstable pay, housing and debt use up the attention people need to retrain. She says AI raises that tax by making schedules less predictable and by adding the strain of supervising AI tools. The research she draws on is solid. A QJE field experiment found that Indian workers paid early were 7% more productive. But no study has measured the AI part of her argument directly. "Cognitive tax" is her own name for the idea, not a new finding.
What changed and event timeline

Cash study shows fading benefits
In OpenResearch's study, people receiving $1,000 a month reported less stress in the first year, but the effect faded by years two and three. Physical and mental health did not improve overall ().
WEF sizes the reskilling need
Employers surveyed by the WEF expect 59 of every 100 workers to need training by 2030. Seven in ten firms rate analytical thinking as essential ().
Walmart expands AI tools for staff
Walmart rolled out AI-powered task and scheduling tools to about 1.5 million US associates ().
BCG names "AI brain fry"
A BCG survey of 1,488 US workers linked heavy supervision of AI tools to mental fatigue, more errors and a stronger wish to quit ().
Brain economy summit in New York
The summit, held during the UN General Assembly, carried the theme "The Human Advantage" and framed brain health as central to the AI transition ().
Moore publishes her proposal
She argues for universal basic services such as free childcare, housing, transport and digital access, and asks the UK to raise "cognitive load" tests during its G20 presidency ().
Capabilities and access
No specific AI model or version is involved.
Read the full section
No specific AI model or version is involved. The AI in question is:
- workforce-management and scheduling software, such as Walmart's associate tools (Walmart);
- general-purpose AI tools and agents that employees have to supervise (The Decoder).
Technical analysis for researchers and developers
- Scarcity meta-analysis: 256 effect sizes, 29 datasets, 111,852 people, with an overall effect of g = −0.43. Education explains about 60% of the effect.
- Causal designs: The India study randomly varied when wages were paid.
- BCG: a cross-sectional survey based on self-reports, so it cannot establish cause.
Read the full section
- Scarcity meta-analysis: 256 effect sizes, 29 datasets, 111,852 people, with an overall effect of g = −0.43. Education explains about 60% of the effect. Long-lasting and severe scarcity does more harm (J. Econ. Psychology 2024).
- Causal designs: The India study randomly varied when wages were paid. Gains came through fewer attention errors and were above 10% for the poorest workers (NBER). The South Africa study used the cut in men's pension age from 65 to 60 as a natural experiment. Men eligible five years earlier were 5.2 percentage points less likely to have dementia (PMC).
- BCG: a cross-sectional survey based on self-reports, so it cannot establish cause.
Claims and evidence
- Scarcity impairs thinking: well supported (meta-analysis). The most direct evidence that the effect lifts when money pressure eases comes from the India experiment (BFI).
- Supervising AI is tiring: workers with heavy oversight duties reported 14% more mental effort and 12% more fatigue (Fortune).
- AI scheduling cuts hours: the evidence is reports from individual workers, such as one person's pay falling nearly 20% (Prism News on NPR).
Read the full section
- Scarcity impairs thinking: well supported (meta-analysis). The most direct evidence that the effect lifts when money pressure eases comes from the India experiment (BFI).
- Supervising AI is tiring: workers with heavy oversight duties reported 14% more mental effort and 12% more fatigue (Fortune). This is BCG's own research, and the sample was workers at large companies, not low-paid workers.
- AI scheduling cuts hours: the evidence is reports from individual workers, such as one person's pay falling nearly 20% (Prism News on NPR). Walmart describes the tools as giving workers more flexibility.
- Insecurity carries an earnings penalty as large as the gender pay gap: this comes from Moore's own institute. No independent confirmation was found.
Context and prior work
Moore builds on the scarcity research of Mullainathan and Shafir, which Kaur et al. extended (MIT). The "brain capital" framing comes from the brain-economy movement, including the Brain Capital Alliance (Rice). Universal basic services is a long-standing proposal from UCL's Institute for Global Prosperity, which Moore founded and runs.
Limitations, safety and contested findings
- No reviewed study measures how AI deployment specifically affects the thinking of low-paid workers.
- It supports Moore's view that cash alone is not enough, and the study's own authors say cash cannot fix childcare or housing costs (Route Fifty).
- Education accounts for most of the scarcity effect (IDEAS), which makes it harder to say how much is caused by insecurity itself.
Read the full section
- No reviewed study measures how AI deployment specifically affects the thinking of low-paid workers. The case for "AI raises the rate" is inferred from separate pieces of evidence.
- The OpenResearch result cuts both ways. It supports Moore's view that cash alone is not enough, and the study's own authors say cash cannot fix childcare or housing costs (Route Fifty). But the reviewed sources contain no trial showing that services work better than cash.
- Education accounts for most of the scarcity effect (IDEAS), which makes it harder to say how much is caused by insecurity itself.
- The piece is opinion from an advocate, so its figures on UK childcare and long-term sickness are not independently checked here.
Business and practitioner implications
- Scheduling systems: if a system optimises only for labour cost, it may push costs onto workers through unstable hours.
- Agent rollouts: BCG reports that perceived productivity peaks at about three AI tools used at once and falls beyond four (The Register).
- Reskilling programmes: the evidence suggests financial stability helps people learn.
Read the full section
- Scheduling systems: if a system optimises only for labour cost, it may push costs onto workers through unstable hours. Firms could track how predictable schedules are as a measure alongside cost.
- Agent rollouts: BCG reports that perceived productivity peaks at about three AI tools used at once and falls beyond four (The Register). Designs could set limits on how much oversight each person carries.
- Reskilling programmes: the evidence suggests financial stability helps people learn. Paying wages early or providing support such as childcare may improve how well training works (NBER).
- Policy: UK G20 priorities could add cognitive-load tests to AI-powered public services.
The source trail.
Sources (17)
AI is imposing a hidden ‘cognitive tax’ on the lowest-paid workers
Article text retrieved; extracted text may omit tables or interactive elements.
theguardian.com