AI-A05 — The Future of AI, Human Agency, and Civilization

Course Code: AI-A05  |  School: School of Artificial Intelligence  |  Cluster: Level 5 — Advanced Systems & Innovation

Level: Advanced  |  Duration: 6 weeks · 24–30 learning hours  |  Language: English  |  Certificate: Professional Certificate (non-degree)  |  Format: Self-paced with AI support under human supervision

Overview

This is the reflective capstone of the School of Artificial Intelligence, and it is a reading and writing course rather than a building course. Its subject is what the diffusion of machine intelligence does to human agency: to the capacity of individuals to form and revise their own judgements, to the capacity of institutions to govern themselves, and to the distribution of power between those who own the systems and those who are processed by them. The question is not whether machines will become conscious. It is what happens to a civilisation that routes an increasing share of its cognition, its attention and its administrative decisions through instruments owned by a small number of firms.

The course is comparative and it is deliberately plural. It reads the technological-determinist accounts alongside their critics, the acceleration literature alongside the political economy of computation, and the Euro-American debate alongside voices from Southeast Asia and the wider global South, for whom the salient questions are dependency, data extraction and the terms on which a technology designed elsewhere is adopted. Learners are not asked to arrive at the instructor’s position; they are asked to arrive at a position they can defend under scrutiny.

Method matters as much as content. Because the subject is saturated with speculation, the course insists on the distinction between a forecast, a scenario, a normative claim and a marketing statement, and requires that learners label their own claims accordingly. The final artefact is a substantial position paper on a question of the learner’s choosing, written to the standard expected of a serious policy or scholarly essay, accompanied by a reflective note on how the learner’s own use of AI shaped the argument.

Learning outcomes

On completion, a successful learner will be able to:

1. Distinguish the principal theoretical positions on technology and society — determinism, social construction, political economy and institutional accounts — and locate a given argument about AI within them.
2. Analyse specific mechanisms by which automated systems alter human agency, including deskilling, automation bias, attention capture, preference formation and the narrowing of the space of available choices.
3. Evaluate claims about labour displacement, productivity and inequality against the available empirical evidence, and identify where a confident public claim rests on speculation.
4. Assess the political economy of contemporary AI: concentration in compute, data and talent, the dependency relations this creates for states outside the producing countries, and the environmental and material costs of the infrastructure.
5. Apply at least two normative frameworks — for example a rights-based framework and a capabilities framework — to a concrete case, and articulate where they diverge in their prescriptions.
6. Construct rigorous scenarios about plausible futures, distinguishing forecast from scenario from advocacy, and stating the indicators that would tell an observer which trajectory is unfolding.
7. Write a sustained, properly referenced argumentative essay that anticipates and answers the strongest objection to its own thesis.

Who this course is for

The course is written for senior professionals and reflective practitioners who must think about AI at the level of institutions and societies rather than tools: policy advisers, university leaders and academics, journalists and editors, civil society and religious institution leaders, strategic analysts, and technologists who want a serious grounding in the debate their work is contributing to. It is also suitable as a standalone course for participants from the humanities and social sciences with no technical background, since no implementation work is required.

Prerequisites

Completion of AI-F01 and AI-F06, or an equivalent working understanding of what contemporary AI systems do and do not do. Learners must be able to read substantial academic texts in English and write extended argumentative prose. No mathematical or programming background is required. A willingness to argue against one’s own initial position is the practical prerequisite.

Syllabus

Module 1 — How to think about technology and society

The opening module supplies the theoretical vocabulary. It sets out technological determinism in its hard and soft forms, the social construction of technology, the political economy tradition, and institutional accounts that treat outcomes as the product of rules and incentives rather than of the artefact itself. Each is applied to the same case so that the differences in explanatory purchase become visible.

Lessons. 1.1 Determinism and its critics. 1.2 Do artefacts have politics? 1.3 Political economy and the ownership of infrastructure. 1.4 Institutions, rules and the shaping of adoption.

The module also establishes the writing standards for the course: claims labelled by type, sources cited to the primary text, and the strongest counter-argument stated in its own best form before being answered.

Deliverable. Analytical note (800 words) applying two frameworks to a single AI case.

Module 2 — Agency, judgement and the cognitive effects of delegation

This module examines what happens to human judgement when it is routinely assisted. It covers automation bias and complacency from the established human-factors literature, the deskilling debate, the effect of recommendation and ranking systems on the formation of preference, and the compression of the choice architecture available to individuals.

Lessons. 2.1 Automation bias and complacency: the evidence. 2.2 Deskilling, upskilling and the ironies of automation. 2.3 Attention, recommendation and preference formation. 2.4 What is lost when the first draft is never yours.

Learners conduct a small self-study of their own delegation practices over one week and analyse what they stopped doing themselves, what capacity they may be losing, and what they gained.

Deliverable. Delegation self-study with analysis (800 words).

Module 3 — Work, value and the distribution of gains

The labour module insists on evidence. It reviews what economic research actually shows about automation and employment, distinguishes task displacement from occupational disappearance, examines the complementarity thesis and its limits, and looks at who captures the productivity gains where they occur. It gives particular attention to the composition of AI labour that remains invisible in the public debate, including data annotation and content moderation work concentrated in the global South.

Lessons. 3.1 Tasks, occupations and the displacement literature. 3.2 Complementarity, the Turing trap and design choices about automation. 3.3 Who captures the gains: wages, profits and market power. 3.4 The hidden labour behind the interface.

Deliverable. Evidence review (1,000 words) assessing one widely repeated claim about AI and employment.

Module 4 — Power, dependency and the material base

This module addresses infrastructure. It covers concentration in compute, foundation models, data and talent; the resulting dependency of states and institutions that consume but do not produce these systems; export controls and the geopolitics of semiconductors; data extraction and the question of who owns the record of a society’s own language and culture; and the energy, water and mineral footprint of the build-out.

Lessons. 4.1 The compute stack and where the chokepoints are. 4.2 Digital dependency and sovereignty for non-producing states. 4.3 Data extraction, language and cultural representation. 4.4 Energy, water and the environmental accounting of AI.

The Southeast Asian case is treated at length: what meaningful technological sovereignty could mean for a middle power, and whether local-language model development, public compute and data-governance regimes are adequate responses.

Deliverable. Dependency analysis of one country or sector (1,000 words).

Module 5 — Normative frameworks and the governance response

Having established mechanisms and interests, the course turns to what ought to be done. It works through rights-based approaches, the capabilities approach, consequentialist risk framings and virtue-oriented accounts of human flourishing, and tests each against hard cases where they yield different answers. It then surveys the governance instruments in circulation — the UNESCO recommendation, risk-tiered regulation, standards-based management systems and voluntary commitments — and assesses their enforceability.

Lessons. 5.1 Rights, capabilities and flourishing. 5.2 Risk framings and their blind spots. 5.3 The instruments: from soft principles to binding rules. 5.4 Enforceability, capture and the limits of self-regulation.

Religious and non-Western ethical traditions are given genuine analytical space rather than a token mention, including Islamic ethical reasoning about knowledge, responsibility and the custodial relationship to creation.

Deliverable. Comparative normative analysis of one hard case (1,000 words).

Module 6 — Scenarios, indicators and the position paper

The final module builds disciplined scenarios. It covers the construction of plausible alternative futures from identified drivers and critical uncertainties, the difference between a scenario and a prediction, the specification of leading indicators that would distinguish trajectories, and the honest treatment of low-probability high-consequence outcomes without either dismissal or sensationalism.

Lessons. 6.1 Drivers, uncertainties and scenario construction. 6.2 Indicators and how you would know you were wrong. 6.3 Catastrophic-risk arguments and how to assess them soberly. 6.4 Writing to persuade a reader who disagrees with you.

The module concludes with the position paper: a sustained argument on a question of the learner’s choosing, with a scenario annex and a reflective note on the learner’s own use of AI in producing it.

Deliverable. Final position paper (3,000–3,500 words) with scenario annex and reflective note.

Assessment

Pass mark 70 per cent. All assessed components must be attempted. Every mark in this course is issued by a human assessor; no assessment outcome is generated automatically.

Rubric criteria

Each assessed artefact is marked against four criteria at four levels (distinction, pass with merit, pass, fail).

1. Conceptual command. Whether the learner uses the theoretical vocabulary accurately, locates arguments within the relevant traditions, and avoids collapsing distinct positions into a single caricature.
2. Evidential discipline. Whether empirical claims are supported by cited primary sources, whether speculation is labelled as such, and whether the learner distinguishes forecast, scenario, normative claim and advocacy.
3. Argumentative rigour. Whether the thesis is clear and contestable, whether the strongest counter-argument is stated in its own best form and answered, and whether the conclusion follows from what precedes it.
4. Situated judgement. Whether the analysis attends to context, power and distributional consequences, including for populations outside the producing countries, and whether recommendations are proportionate and actionable.

Reading list

Core. Kate Crawford, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (New Haven: Yale University Press, 2021).

Langdon Winner, ‘Do Artifacts Have Politics?’, Daedalus, 109:1 (1980), pp. 121–136.

Shoshana Zuboff, The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power (New York: PublicAffairs, 2019).

Ruha Benjamin, Race After Technology: Abolitionist Tools for the New Jim Code (Cambridge: Polity Press, 2019).

Erik Brynjolfsson & Andrew McAfee, The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies (New York: W. W. Norton, 2014).

Nick Bostrom, Superintelligence: Paths, Dangers, Strategies (Oxford: Oxford University Press, 2014).

UNESCO, Recommendation on the Ethics of Artificial Intelligence (Paris: UNESCO, 2021).

Academic integrity and use of AI

This course asks learners to think for themselves about what it means to delegate thinking, and the assessment is designed accordingly. AI tools may be used for search, summarisation of sources the learner has independently obtained, language editing and criticism of drafts. They may not be used to generate the argument itself. Every submission carries a use-of-AI statement specifying what was used and for what; the final position paper additionally requires a reflective note on how that use shaped the argument, and this note is marked.

Every citation must be verified against the primary source and must exist. Fabricated references and invented quotations are treated as the gravest breach in this course, and any submission containing them will be failed and referred. Learners are reminded that a fluent paragraph produced by a model that they cannot explain, defend or trace to a source is not their work in any sense the certificate can vouch for.