Is AI Going to Have a Dehumanising Effect on Humans?, An Extended Analysis

Abstract

This paper asks whether artificial intelligence (AI) will dehumanise human society. Its central claim is that AI is not, by itself, a dehumanising force: whether it diminishes or elevates us depends on how we design, regulate, and integrate it. To test that claim, the paper looks at three earlier technological revolutions, industrial machinery, the automobile, and the ATM, and asks whether their reassuring "the jobs came back" pattern is guaranteed to repeat with AI. It argues that it is not guaranteed: recent economics shows that some automation is "so-so", it displaces workers without producing the broad gains that historically created new work (Acemoglu & Restrepo, 2019). The paper therefore keeps its optimistic thesis but makes it conditional: AI will augment rather than replace human potential only if we build the right guardrails. It closes with three concrete policy mechanisms, risk-tiered regulation with content-labelling, portable reskilling accounts, and wage insurance, that would tilt the outcome toward the humanising side.

Introduction

AI has evolved from simple automation into complex systems that can process information faster than any individual. Opinion is split: supporters see enormous efficiency gains, while critics fear a loss of human value, judgement, and emotional connection. This paper examines the history of technology, the current state of AI, and its psychological and economic effects to answer one question: will the relationship between humans and machines be adversarial or collaborative?

The argument is this: technology does not shape society on its own; society shapes how technology is used. AI will only dehumanise us if we let it, and avoiding that outcome is a matter of deliberate design and policy, not fate. To make that case honestly, the paper does something most optimistic accounts skip: it engages the strongest evidence that AI might break the historical pattern, and then shows what it would take for the optimistic outcome to actually happen.

Historical Perspective: Learning from Past Technological Revolutions

The Industrial Revolution

In the late 1700s and 1800s, workers feared that mechanised looms and steam engines would erase their livelihoods, and the Luddite movement smashed the new machines in protest. Yet industrialisation ultimately created millions of new positions, in factory work, engineering, railways, mining, textile production, and construction, along with entirely new professions such as mechanical engineers and industrial managers.

The Automobile (early 1900s)

Early fears centred on the loss of horse-related work, stable-hands, farriers, carriage-makers. Instead, the automobile grew into one of the largest industries in the world, employing people in vehicle manufacturing, steel, oil refining, tyre production, road building, maintenance, and public transport. Today the global auto industry supports roughly 5% of the world's manufacturing employment and remains one of the planet's largest single employers (OICA). What matters for our question is the pattern, not a precise headcount: a technology that destroyed some jobs went on to anchor whole new sectors.

The ATM (1967 onwards)

When automated teller machines arrived, banks and tellers feared mass job losses. The most careful study of what actually happened found the opposite. As economist James Bessen documented, the number of tellers needed to run the average urban branch fell from about 20 to 13 between 1988 and 2004, but because ATMs made branches cheaper to operate, banks opened far more of them (urban branches rose about 43%). The result: even as more than 400,000 ATMs were installed across the United States, the total number of teller jobs did not fall (Bessen, IMF 2015). Tellers also shifted from routine cash-handling toward relationship and advisory work.

What History Suggests, and What It Doesn't Guarantee

The lesson usually drawn is comforting: new technology hurts labour in the short term by eliminating obsolete tasks, then creates new, often higher-skilled and better-paid work. Fear is normal at first, but societies adapt and benefit.

That lesson is real, but it is not a law of nature. In each case above, new work appeared because the technology raised productivity enough to expand demand and because new tasks were reinstated for humans to do. As the next section argues, neither of those conditions is automatic, and there are reasons to think AI may not repeat the pattern as smoothly as the looms-cars-ATMs story implies.

Why AI Might Be Different, Engaging the Counter-Case

It is easy to answer every automation worry with "we've heard this before." But the honest version of this paper has to take the opposing evidence seriously, because a growing body of research suggests that how automation happens matters as much as whether it happens.

Economists Daron Acemoglu and Pascual Restrepo describe automation as a tug-of-war between two forces: a displacement effect, which removes human labour from tasks, and a productivity (reinstatement) effect, which lowers costs and creates new tasks and jobs elsewhere (Acemoglu & Restrepo, 2019). Historically the second force eventually won. But they warn about a category they call "so-so technologies": automation that displaces workers without generating much of a productivity boost, think self-checkout kiosks or automated phone menus, so it removes jobs while creating too little new value to replace them (MIT Sloan, 2020).

The empirical record backs the caution. Studying US industrial robots, Acemoglu and Restrepo found that each additional robot per thousand workers reduced the local employment-to-population ratio by roughly 0.18 to 0.34 percentage points and wages by 0.25 to 0.5 percent, with no offsetting boom nearby (Acemoglu & Restrepo, 2020). Separately, they estimate that the automation of routine tasks explains 50 to 70% of the rise in US wage inequality between 1980 and 2016, with real wages for men without a high-school diploma falling about 8.8% over that period (Acemoglu & Restrepo, 2022). In other words, recent automation has already polarised the labour market, hollowing out middle-skill work, rather than simply lifting everyone into better jobs.

Even mainstream forecasting is more cautious than the popular "AI will create tens of millions of net new jobs" line suggests. The World Economic Forum's Future of Jobs Report 2023 projected that, over 2023 to 2027, employers expected 69 million new jobs to be created but 83 million eliminated, a net loss of about 14 million jobs, or 2% of employment (WEF, 2023). The long-run picture may still turn positive, but the near-term evidence points to net displacement, not an automatic jobs bonanza.

This does not overturn the paper's thesis; it sharpens it. History shows the humanising outcome is possible, not inevitable. Whether AI reinstates enough new human work, or ends up as economy-wide "so-so" automation that displaces without replacing, depends heavily on the choices we make. That is precisely why policy, addressed later, is the hinge of the whole argument.

Why People Fear AI More Than Previous Technologies

Even setting economics aside, AI provokes disproportionate anxiety. The specific fears include total labour replacement, the loss of creativity, machines mimicking emotion, unhealthy dependency, erosion of privacy, and machines making critical decisions on their own.

AI Learns and "Thinks"

Unlike earlier machines that performed physical labour through rigid, pre-set automation, AI reaches into cognitive work: data analysis, pattern recognition, prediction, problem-solving, creative output (drafting text, composing music, generating images), and context-aware conversation. Because it mimics thinking, AI can feel like a direct competitor to human intelligence rather than a mere tool.

A Personal, Human-Like Connection

Conversational AI, humanoid robots, and generative images resemble human abilities closely enough to trigger the "uncanny valley", an unease we simply don't feel toward a washing machine or a tractor.

An Unprecedented Rate of Change

AI has advanced far faster than past technologies. Generative capabilities matured in a few years, whereas cars and computers took decades to become ubiquitous, leaving less time for institutions and workers to adjust.

Media Amplification

Films, sensational news, and viral posts inflate fear with headlines like "AI will take your job" and "Robots will take over the world." Anxiety rises further when respected experts publicly disagree about where AI is heading.

Will AI Dehumanise Humans? Understanding the Concept

"Dehumanisation" here means five things: job loss and a loss of purpose; reduced creativity and thinking skills; less human interaction; unhealthy dependence on machines; and weakened emotional intelligence and ethics. The paper takes each in turn.

Jobs and Purpose: Will AI Replace Human Work?

AI will certainly take over repetitive, rule-based tasks: basic data entry, routine bookkeeping, script-based customer service, and standardised manufacturing steps.

Short term: disruption is real. As the evidence above shows, low-skill, routine roles face elimination or restructuring, and the transition can depress wages and widen inequality if left unmanaged (Acemoglu & Restrepo, 2022). This is not a small footnote, it is the part of the story where real people are hurt, and where policy has to do the work.

Long term: the argument is that labour shifts toward higher-value roles that draw on distinctly human strengths, complex critical thinking, abstract creativity, high emotional intelligence, and nuanced, context-heavy judgement, which AI cannot fully replicate. New specialised jobs are already emerging: AI developers and architects, machine-learning engineers, robotics technicians, AI-assisted healthcare diagnosticians, AI ethics auditors and governance leads, cybersecurity professionals, and prompt and content specialists. But the long-term optimistic outcome is conditional on the productivity and reinstatement effects winning out, which, as the WEF's near-term net-negative projection reminds us, is not guaranteed and must be actively supported (WEF, 2023).

Creativity: Will AI Shrink Human Imagination?

History is reassuring here. When cameras democratised image-making and digital synthesizers reshaped music, purists claimed technology had "killed art", yet art evolved into new forms rather than dying. AI offers similar tools: designers move from physical sketches to computer-assisted layout; writers use AI to brainstorm and refine drafts; musicians reach for AI-generated sounds and virtual instruments; filmmakers use AI storyboarding to visualise a scene before spending money to shoot it. Because humans supply the lived experience, emotion, and intent, AI acts as powerful assistance rather than a replacement. Human creativity is not destroyed; it is enhanced and democratised.

Human Interaction: Will AI Make Us Less Social?

Heavy use of chatbots, virtual assistants, and social robots raises real concerns. But AI does not inherently reduce human contact, people retain agency over when to choose a person over a machine. The telephone was once criticised as a threat to face-to-face interaction, yet it became one of the most effective tools ever for maintaining relationships across distance. Similarly, AI can strengthen connection through real-time translation, wider access to high-quality education, links to isolated rural communities, and smoother collaborative work. Whether technology isolates or connects us depends on our choices, not on the technology itself.

Dependence on AI for Decision-Making

AI will increasingly assist decisions in medicine, logistics, transport and aviation, agriculture, and finance. But humans must stay in charge through "human-in-the-loop" (HITL) oversight, because AI lacks moral judgement: it does not grasp the philosophical weight of right versus wrong, cannot truly feel or respond to complex emotions, cannot be held legally or morally accountable, and follows programmed responses rather than genuine ethical intuition. AI can be the ultimate data-processing co-pilot, but the human must remain the final, authoritative decision-maker.

Emotional and Ethical Concerns

AI has no genuine empathy, feels no pain, and cannot form real bonds of love, kinship, or friendship. The risk of emotional detachment comes not from AI itself, which is only code and cannot generate malice on its own, but from people misusing it as a crutch for real social interaction. Clear, widely shared ethical standards and enforceable regulation can reduce these harms.

The Positive, Humanising Effects of AI

By automating drudgery and freeing people for creative, caring, and strategic work, AI could make society more humane and more focused on well-being.

Eliminating Drudgery

Automating repetitive, mind-numbing tasks can liberate humans for creativity and the arts, scientific innovation, caregiving and emotional labour, and high-level strategy.

Transforming Healthcare

AI supports clinicians with faster early diagnosis (such as cancer detection in radiology), personalised and genetically informed treatment, accelerated medical research, and predictive models that help prevent outbreaks, improving quality of life, longevity, and dignity.

Democratising Education

AI can provide accessible, 24/7 tutoring and adaptive learning that lets students progress at their own pace, helping to level the playing field for disadvantaged communities.

Improving Accessibility

For people with disabilities, AI enables real-time speech recognition and generation, live image description for the visually impaired, assistive mobility robots, and personalised devices that adapt to neurodivergent needs, technologies that actively uplift and protect society.

What Would Actually Keep AI Humanising: Three Policy Mechanisms

If the humanising outcome is a choice rather than a certainty, then vague calls for an "appropriate regulatory framework" and "ethical development" are not enough. The choice has to be built into institutions. Three concrete, already-tested mechanisms would do most of the work.

1. Risk-tiered regulation with mandatory labelling of AI content. The European Union's AI Act, the world's first comprehensive AI law, sorts systems into four tiers, unacceptable (banned outright, e.g. social scoring), high-risk (strict duties: risk management, human oversight, documentation, conformity assessment), limited-risk (transparency: users must be told they are dealing with AI), and minimal-risk (no special rules) (EU AI Act, high-level summary). Crucially, Article 50 requires that AI-generated audio, image, video, and text be marked in a machine-readable format as artificially generated, and that deepfakes be disclosed to viewers, obligations that become enforceable on 2 August 2026 (EU AI Act, Article 50). Proportionate rules plus mandatory "made by AI" labels directly address the fears of deception, manipulation, and unaccountable decision-making the paper identifies.

2. Portable, publicly funded reskilling accounts. The paper repeatedly relies on workers "reskilling," but reskilling only happens if it is funded and accessible. Singapore's SkillsFuture programme shows how: every citizen receives an individual, non-expiring training credit they control directly, topped up with an additional S$4,000 for mid-career workers aged 40 and above (from May 2024), plus a training allowance covering up to 50% of average income (capped at S$3,000/month) for full-time courses (Singapore MOE, 2024). Because the credit is portable and attached to the person rather than the employer, it follows workers through exactly the transitions AI will force.

3. Wage insurance to cushion displaced workers. Even with retraining, many displaced workers land in lower-paying jobs. Wage insurance tops up part of the gap between old and new pay, making re-employment worthwhile. The United States already runs a version, Reemployment Trade Adjustment Assistance, which pays workers aged 50 and over up to half the difference between their old and new wages for up to two years (New York Fed, 2024). Recent research found that this wage insurance raised workers' long-run earnings and was essentially self-financing, because higher employment and tax receipts offset its cost (Hyman, Kovak & Leive, 2024). Scaling such a scheme beyond trade to technological displacement would blunt the inequality that "so-so" automation tends to produce.

Together, these three mechanisms convert the paper's hope, that society will steer AI well, into specific, fundable institutions: rules that make AI honest, credits that make reskilling real, and insurance that makes the transition survivable.

Future Outlook: 2030 to 2035 and Beyond

Many analysts expect that between 2030 and 2035 AI will reach broad maturity and integrate across global industries. Handled well, that era could balance early job losses with new job creation, build entirely new employment ecosystems, deepen human-AI collaboration, raise productivity, and demand reformed education and new skills. The overarching goal should be augmentation, not replacement, but the evidence in this paper is a reminder that "handled well" is doing all the work in that sentence. The augmentation outcome is a target to be secured through policy, not a default to be assumed.

Conclusion

Like the steam engine, electricity, and the internet before it, AI brings both disruption and opportunity. History shows that societies can absorb such shocks and emerge better off, but it also shows, through the economics of "so-so" automation and rising wage polarisation, that good outcomes are earned, not guaranteed (Acemoglu & Restrepo, 2019; 2022). The near-term data are sobering: the WEF's own employers expect a net loss of jobs before any long-run gains arrive (WEF, 2023).

That is why the thesis holds with a condition attached. AI will not dehumanise people unless we build systems that let it, and it will humanise us only if we deliberately build systems that make it. With risk-tiered regulation and honest content labelling, portable reskilling accounts, wage insurance for the displaced, broad public education, and transparent debate about both the benefits and the harms, people can use AI to reach their fullest potential. Technology does not shape society on its own; society shapes how technology is used. Approached that way, AI is best understood not as an existential threat, but as an expression of human ingenuity, creativity, and ambition, one whose humanising promise we still have to choose.

Sources

1. Acemoglu, D. & Restrepo, P. (2019). "Automation and New Tasks: How Technology Displaces and Reinstates Labor," Journal of Economic Perspectives (NBER Working Paper 25684). https://www.nber.org/papers/w25684
2. MIT Sloan (2020). "The lure of 'so-so technology,' and how to avoid it." https://mitsloan.mit.edu/ideas-made-to-matter/lure-so-so-technology-and-how-to-avoid-it
3. Acemoglu, D. & Restrepo, P. (2020). "Robots and Jobs: Evidence from US Labor Markets," Journal of Political Economy (NBER Working Paper 23285), one robot per thousand workers cut the employment-to-population ratio by 0.18 to 0.34 pp and wages by 0.25 to 0.5%. https://www.nber.org/papers/w23285
4. Acemoglu, D. & Restrepo, P. (2022). "Tasks, Automation, and the Rise in US Wage Inequality," Econometrica, automation explains 50 to 70% of the 1980 to 2016 change in the US wage structure (MIT/EurekAlert summary). https://www.eurekalert.org/news-releases/972022
5. World Economic Forum (2023). Future of Jobs Report 2023, 69 million jobs created, 83 million eliminated, a net loss of 14 million (2% of employment), 2023 to 2027. https://www.weforum.org/press/2023/04/future-of-jobs-report-2023-up-to-a-quarter-of-jobs-expected-to-change-in-next-five-years/
6. Bessen, J. (2015). "Toil and Technology," IMF Finance & Development, ATMs and bank tellers: tellers per branch fell from 20 to 13 (1988 to 2004) while urban branches rose 43%, so total teller jobs did not fall. https://www.imf.org/external/pubs/ft/fandd/2015/03/bessen.htm
7. EU AI Act, High-level summary of the four risk tiers. https://artificialintelligenceact.eu/high-level-summary/
8. EU AI Act, Article 50 (transparency obligations; machine-readable marking of AI-generated content; deepfake disclosure). https://artificialintelligenceact.eu/article/50/
9. Singapore Ministry of Education (2024). "New SkillsFuture Level-Up Programme", non-expiring S$4,000 mid-career credit and training allowance. https://www.moe.gov.sg/news/press-releases/20240304-new-skillsfuture-level-up-programme-to-boost-mid-career-upskilling-and-reskilling
10. Federal Reserve Bank of New York (2024). "Wage Insurance: A Potential Policy for Displaced Workers," Liberty Street Economics, Reemployment Trade Adjustment Assistance design (age 50+, up to half the wage loss, up to two years). https://libertystreeteconomics.newyorkfed.org/2024/07/wage-insurance-a-potential-policy-for-displaced-workers
11. Hyman, B., Kovak, B. & Leive, A. (2024). "Wage Insurance for Displaced Workers," NBER Working Paper 32464, wage insurance raised long-run earnings and was self-financing under conservative assumptions. https://www.nber.org/papers/w32464