DR PETER FIELDS • AI & IDEAS

THE AI NOTEBOOK / WHAT MATTERS AND WHY

The Luddite Instinct

Dr Peter Fields •


Download this article (PDF)

AI Explained · Why We Fear AI, Part 1

By Dr Peter Fields

THE 2-MINUTE SUMMARY

Resistance to AI at work can be about income, pride in a craft or having a say in how the job is done. The Luddites knew their machines well. The terms on which those machines were used mattered to them.

Generative AI can take over parts of a job. There are signs of pressure on some entry-level roles, even though broader employment data show no clear evidence of widespread displacement.

Employers who automate junior work should explain where the next generation will learn the judgement that work once taught.

What the Luddites were defending

In 1811 framework knitters in Nottinghamshire began breaking stocking frames. These were skilled hosiery workers who knew the machinery. They objected to practices that included using wide frames to make cheaper stockings at lower rates of pay. In 1812 the disturbances spread to Yorkshire, where cloth finishers attacked shearing frames, and to Lancashire, where weavers attacked power looms. Restrictions on workers organising made it harder to bargain over the changes.1

Calling a colleague a Luddite can end a conversation that ought to begin with a question. What do they stand to lose? An illustrator may be worried about this month’s commissions. A junior lawyer may wonder how to learn a craft if the early work goes to software. A customer adviser may fear losing discretion over how to help someone. Each concern calls for a different response.

What people say they fear

In the UK’s 2025/26 public engagement survey, 74% of adults identified inaccurate information as a risk of AI use. Privacy and security were selected by 69%, job losses by 63%. Respondents could identify several risks. The figures describe what people are worried about. They cannot tell us whether those harms will occur.2

This series looks at work, creativity, education, infrastructure and the risk of losing control of a system. Privacy and surveillance, convincing fakes, bias, dependence on AI companions and concentrated corporate power deserve attention in their own right. Other articles in the series discuss Confidentiality and Privacy, Deepfakes and whether AI can offer the kind of connection we find in another person.

Geek Box · Why control matters

Slovic’s research on risk perception helps explain why a hazard can feel different when you choose it and when someone imposes it on you. Kunda’s work on motivated reasoning offers another clue. What we want to believe can affect which evidence comes to mind and how we weigh it. That applies to the manager who has championed a new system as much as to someone whose job depends on the existing way of working. A worker can also be right about a risk to their livelihood.3,4

A job is made up of tasks

A tool can extract dates from a bundle and draft a chronology. Someone still needs to decide which dates matter, notice a conflict between the records and a witness’s account, and take responsibility for the advice. Automating the first task changes the job. Whether it removes a post depends on what the employer does next.

The International Labour Organization estimates that roughly one job in four worldwide is in an occupation with some exposure to generative AI. About 3.3% of employment falls into its highest exposure category. That highest share was 4.7% for women and 2.4% for men, in part because clerical work is particularly exposed. The study classifies what a tool could do. Its authors regard changes to jobs as the more likely outcome and do not count actual redundancies.5

The labour-market evidence gives a more immediate reason to pay attention. In August 2026, Brynjolfsson, Chandar and Chen reported that employment of US workers aged 22–25 in exposed occupations was about 19% below where it would have been had it kept pace with less-exposed peers. Experienced workers showed no comparable gap. In the UK, a DSIT and LinkedIn snapshot found entry-level hiring for graphic designers down 28% year on year, while entry-level hiring overall broadly followed the wider market. These are patterns in different measures, and the studies do not establish AI as the cause. Yale’s broader US figures give little evidence of widespread disruption so far.6,7,8

How will the next person learn?

Autor argues that AI could help people with relevant knowledge take on tasks previously reserved for a narrower group of experts. That is a possibility worth testing. In a law firm, I would give a junior the machine’s chronology and the original records, and ask them to explain what it got wrong and which facts change the advice. That preserves some of the work of learning while saving some of the time. Responsibility cannot be delegated to a probability distribution.9

I would ask an employer planning to automate junior work what it expects the new recruits to learn and how it will know they have learnt it. The problem comes into sharper focus in creative work, where the skill being copied may be part of what a person sells. The related series Who Will Become the Experts? takes that question further.

Sources and notes

1 The National Archives, Why did the Luddites protest? Original documents on the Nottinghamshire disputes and background on the spread of Luddism.

2 Department for Science, Innovation and Technology, Public Engagement Survey 2025/2026, chapter 6.

3 Paul Slovic (1987), Perception of Risk, Science 236, 280–285.

4 Ziva Kunda (1990), The Case for Motivated Reasoning, Psychological Bulletin 108(3), 480–498.

5 Pawel Gmyrek, Janine Berg, Karol Kamiński, Filip Konopczyński, Agnieszka Ładna, Balint Nafradi, Konrad Rosłaniec and Marek Troszyński (2025), Generative AI and Jobs: A Refined Global Index of Occupational Exposure, ILO Working Paper 140.

6 Erik Brynjolfsson, Bharat Chandar and Ruyu Chen (2025, revised August 2026), Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence.

7 Department for Science, Innovation and Technology and LinkedIn (8 June 2026), A Snapshot of Entry-Level Hiring in the UK, Figure 2.

8 Yale Budget Lab, AI Labor Market Tracker, September 2026 data update.

9 David Autor (2024), Applying AI to Rebuild Middle Class Jobs, NBER Working Paper 32140.

Download this article (PDF)