Sharing automation’s gains matters. So does preserving the knowledge, resources and authority to decide what automation is for.
My illustration entitled: “The Loom of the Automated Future” – intelligent machines weave a vast tapestry depicting tomorrow’s economy. People hold the design patterns, examine alternative versions and jointly redirect the emerging weave. The future should remain open to human revision as automation advances.
The value of a dividend, and the question it leaves open
I welcome the possibility that automation could help people work less dangerously, obtain better services and enjoy greater economic security. If artificial intelligence raises productivity, sharing its gains through higher wages, shorter hours, public services, profit-sharing or income support could improve lives. A financial dividend would be a substantial achievement where the alternative is concentrated wealth and widespread insecurity.
But imagine receiving a regular payment from an automated economy whose direction you cannot meaningfully question. You receive some proceeds, yet cannot influence which work is redesigned, which services become impersonal or which activities are measured chiefly by speed. Distribution answers who benefits from a settlement. It does not necessarily answer who can reconsider the settlement itself.
These questions overlap. Secure income can give people time to organise and bargaining power to refuse harmful work. Ownership can bring voting rights. Consultation can change a deployment. My concern is whether those possibilities become dependable capacities, rather than benefits granted on terms that others can unilaterally withdraw.
Reauthorism proposes that essential economic arrangements sustain affected publics’ independently governed knowledge, resources and practical capacity to reconsider their purposes, rules and structures. Essential arrangements are systems whose operation or exclusion can seriously affect basic living conditions and economic participation, particularly where alternatives are weak. The obligation should be proportionate to that dependence, rather than extending unrestricted public control over every business.
What the evidence available in 2023 could establish
By August 2023, there were reasons to take AI’s economic possibilities seriously without assuming a fully automated economy was imminent. The April 2023 version of Erik Brynjolfsson, Danielle Li and Lindsey Raymond’s Generative AI at Work studied a conversational assistant used by customer-support agents. It reported productivity gains that were greatest among less experienced and lower-skilled workers. This was evidence about assistance in a particular setting, not proof that whole occupations could be replaced or that gains would automatically reach everyone. [1]
That distinction matters politically. A technology assisting workers presents different choices from one eliminating their discretion. Even when a tool performs a task successfully, institutions must still decide how it fits into employment, service delivery and responsibility for mistakes. Technical capability cannot determine those purposes by itself.
I do not need to predict universal technological unemployment to ask who shapes these choices. A system that changes the pace, supervision or conditions of existing work already changes economic life. The governance question arises before any speculative future in which machines perform nearly all productive activity.
Income, ownership, consultation and authorship
Income support provides purchasing power and security. In Basic Income, Philippe Van Parijs and Yannick Vanderborght defend an unconditional income within a broader argument about freedom and economic insecurity. That debate should be engaged on its own terms, rather than reduced to paying people to accept powerlessness. [2] Nevertheless, even generous income cannot alone specify who may alter an automated service’s mandate.
Ownership of productive assets offers another route. Shares, collective funds or worker ownership can distribute returns and governance rights. But a diversified investor may have little practical influence over deployment; employee owners may disagree with users or communities. Ownership matters without automatically representing every person substantially affected.
Robert Dahl’s defence of economic democracy connected control of economic resources with political equality. Reauthorism inherits that concern rather than discovering it. [3] Its question here is whether people retain usable authority to revisit an essential arrangement’s purposes, including under ownership forms established through reform.
Consultation, meanwhile, offers an opportunity to be heard. Authorship requires an effective relationship between participation and consequential decisions: the ability to put alternatives on the agenda, obtain independent examination and use defined procedures to seek change. It need not give everyone control of daily operations or an unlimited veto. It must make more possible than commenting on an outcome already decided.
A hypothetical housing-repairs service
Consider a hypothetical municipal housing authority procuring AI tools to classify written repair requests and draft replies for staff. Officials expect quicker responses and lower administrative costs. They propose sharing verified savings through tenant rent credits and a staff bonus. Those benefits deserve consideration, alongside the proposal’s costs and uncertainties.
The authority initially measures success by requests processed and cases closed. Workers worry that difficult cases will become performance liabilities. Tenants want fewer delays but fear that an incorrectly classified report of damp or unsafe wiring will receive a reassuring reply instead of an inspection. The ability to generate fluent text does not establish that a building is safe.
Affected non-users also have interests. Neighbours can face consequences from unresolved hazards; people awaiting public housing have claims on the same limited budget. Contractors and municipal staff must carry out decisions. The supplier knows its product, but the authority remains responsible for choosing and overseeing the service.
A Reauthorist process would begin before procurement fixes the available choices. Worker representatives, tenant representatives and appropriately selected affected non-users would receive funded access to independent technical and housing expertise. They could examine whether the real objective should be faster correspondence, faster repairs, better identification of danger or improved access for people who struggle with written forms.
Those purposes suggest different designs. A limited pilot might test classification while retaining human assessment of urgency. Staff might spend saved administrative time assisting vulnerable tenants. The authority could decide that safety-critical prioritisation should remain outside automation, while routine drafting proceeds. Deliberation might also support the original proposal after adequate examination. Authorship does not require rejecting technology.
Decision-making responsibilities must be clear. The elected authority would approve service priorities and budgets; worker representatives would negotiate applicable employment arrangements; a representative review body could commission evaluations and require reasoned responses. Users would retain accessible routes to challenge errors. These are proposed allocations of authority, subject to the jurisdiction’s lawful institutional design.
My illustration “The Loom of the Automated Future” work-in-progress. The art represents that sharing automation’s wealth is insufficient unless people retain the knowledge, tools and authority to revise the economic future it weaves.
Understanding enough to exercise authority
No sensible system requires every tenant or worker to become a machine-learning specialist. People can judge purposes and consequences while relying on qualified assistance, just as elected representatives rely on engineering advice without personally designing bridges. What matters is whether they can obtain advice independent of those promoting the deployment and understand its practical implications.
The OECD’s July 2023 Employment Outlook already examined social dialogue around AI, including access to technical expertise. It reported associations between consultation and more positive assessments of AI’s effects, while explicitly warning that the evidence did not establish causation. [4] Participation and supported technical understanding were therefore established concerns by the reference date.
A Reauthorial Dividend would support this capacity rather than necessarily pay cash. In the housing example, a proportionate allocation could fund independent evaluation, paid participation, accessible explanations and preparation for possible changes. Funding should not depend entirely on savings materialising: people need capacity to assess a proposal before its financial benefits exist.
Experts should disclose relevant interests, explain uncertainties and distinguish technical constraints from policy preferences. Competing analysis may be justified where consequences are substantial. Public participants need enough time and support to question recommendations. An expert’s ability to describe a model does not give that expert an exclusive claim to decide what the housing service should value.
Choosing purposes before choosing suppliers
Procurement can quietly settle political questions. If a contract rewards rapid closure of requests, replacing the supplier may preserve the same disputed objective. Public authorship must therefore reach the specification of the service, its performance criteria and the conditions under which its purpose can be reconsidered.
The authority should seek usable documentation, access to relevant performance evidence, proportionate audit provisions and workable transition assistance. It should retain staff who understand the service and can assess alternative designs. Records and necessary capabilities should survive supplier changes, so a successful deployment does not gradually become an arrangement nobody outside the supplier can understand.
Review should continue at reasonable intervals, with earlier examination where credible evidence of serious harm arises. A scheduled review would ask whether the service’s goals remain appropriate, not merely whether the software meets its contract. Changes could be staged, with fallback arrangements and explicit budgets for retraining or transition. Continued use should remain a legitimate outcome. The point is to preserve a practical choice while managing the costs that changing direction imposes on workers, users and suppliers.
This is not an unprecedented demand for human alternatives. The White House’s October 2022 Blueprint for an AI Bill of Rights proposed consultation, independent evaluation, human alternatives and the possibility of not deploying or removing unsafe or ineffective systems. It offered guidance, not a general statutory entitlement. [5] Reauthorism’s additional organising concern is preserving materially supported reconsideration even when a deployment performs successfully against its existing targets.
The costs and conflicts cannot be wished away
The strongest practical objection is that governance can slow useful innovation. Suppliers need returns on investment and predictable contracts. A small developer cannot finance unlimited bespoke disclosures or repeated evaluations. International competition may penalise institutions that take longer to introduce effective tools. These costs belong inside the argument.
I would favour proportionality: deeper review where systems affect essential needs, impose serious harms or create durable dependence; lighter procedures for limited, reversible uses. Shared evaluation resources, standard documentation and scheduled reviews can reduce duplication. Defined decision periods and staged commitments can give suppliers reasonable certainty while preserving routes to reconsider consequential choices.
Commercial confidentiality also has legitimate purposes. Meaningful scrutiny need not require publishing source code, model weights or every trade secret. Trusted independent evaluators can examine restricted material under appropriate safeguards, while publishing findings and reasons understandable to affected people. But confidentiality cannot justify withholding all evidence needed to assess a service. If adequate scrutiny is impossible, that limitation should affect the procurement decision.
Publics will disagree. Workers may prefer employment guarantees; tenants may prefer lower rents; applicants waiting for housing may favour spending savings on additional homes. Representation should recognise these competing interests without pretending that any group speaks for everyone. Reasons, minority reports, appeal routes and clearly allocated final authority can make disagreement manageable without manufacturing consensus.
Participation itself can privilege organised interests and people with spare time. Archon Fung’s work distinguishes who participates, how discussion operates and how it connects to public action. [6] Selection procedures, compensation, accessible information and scrutiny of the participation fund matter as much as the invitation to attend.
A dividend that preserves the next decision
Debates about technological unemployment ask how work and livelihoods may change. Basic-income debates consider security and distribution. Economic democracy examines control over productive institutions, while participatory governance explores how people influence collective decisions. Reauthorism overlaps with all of them. Its proposed contribution is a continuing duty to preserve independent revision capacity within essential arrangements, including prosperous ones.
Financial benefits can help create that capacity, and institutions should not demand that people choose between income security and meaningful voice. The question is whether a share of prosperity also supports the knowledge, representation and practical alternatives needed to reconsider how prosperity is produced.
I would judge an AI dividend partly by the next decision it leaves open. Can affected people question the objectives, examine alternatives, change deployment conditions or decide that an activity should remain human? An automated economy becomes more answerable when its beneficiaries possess the supported authority to help write its next stage.
Sources
- Erik Brynjolfsson, Danielle Li and Lindsey Raymond, Generative AI at Work. April 2023 working paper; linked version submitted 23 April 2023. The article uses this historical version rather than subsequent revisions.
- Philippe Van Parijs and Yannick Vanderborght, Basic Income: A Radical Proposal for a Free Society and a Sane Economy. Harvard University Press, 2017; linked author-institution publication record.
- Robert A. Dahl, A Preface to Economic Democracy. University of California Press, 1985; linked paperback edition published August 1986.
- OECD, “Social dialogue and collective bargaining in the age of artificial intelligence”, Chapter 7 of OECD Employment Outlook 2023: Artificial Intelligence and the Labour Market. Published 11 July 2023.
- White House Office of Science and Technology Policy, Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People. October 2022; linked official archived material.
- Archon Fung, “Varieties of Participation in Complex Governance”. Public Administration Review 66(s1), December 2006, pp. 66–75; first published online 9 November 2006.