Recalibrating assumptions on AI

Towards an evidence-based and inclusive AI policy discourse

Research paper

Published 12 April 2023

ISBN: 978 1 78413 562 1

Arthur Holland Michel

Carnegie Council Senior Fellow

Many assumptions about artificial intelligence (AI) have become entrenched despite the lack of evidence to support them. Basing policies on these assumptions is likely to increase the risk of negative impacts for certain demographic groups. These dominant assumptions include claims that AI is ‘intelligent’ and ‘ethical’, that more data means better AI, and that AI development is a ‘race’.

The risks of this approach to AI policymaking are often ignored, while the potential positive impacts of AI tend to be overblown. By illustrating how a more evidence-based, inclusive discourse can improve policy outcomes, this paper makes the case for recalibrating the conversation around AI policymaking.

DOI: 10.55317/9781784135621

Summary

Artificial intelligence (AI) policy is underpinned by a range of common assumptions about how AI will contribute to economic, military and societal advantage, how such ‘AI power’ can be harnessed, and how the technology’s known risks can be averted. Many of these assumptions have become entrenched despite the fact that they do not represent the interests of all stakeholders and they misrepresent a growing body of evidence. Policies built upon these assumptions could lead to elevated risks for certain demographic groups, among other negative outcomes.
As AI policies begin to take the form of hard rules and regulations, the common assumptions that underpin them must be able to accommodate new facts and be comprehensive enough to account for all possible risks. Most importantly, they must be representative of the interests of all stakeholders affected by the technology and the norms that govern it.
Examples of dominant AI assumptions include the claim that AI is ‘intelligent’, that ‘more data’ is a requisite for better AI, that AI development is ‘a race’ among states, and that AI itself can be ‘ethical’.
These types of assumption are unyielding to contrary evidence, counterfactuals and nuance that challenges their standing. They extol the virtues of AI but neglect to consider its failures, while tending to serve a narrow but powerful set of interests. And they minimize perspectives from beyond the Global North and the male-dominated tech sector, as well as perspectives centred around non-technical approaches to societal problems.
The dominance of these AI assumptions is problematic. The greater the gap between a policy assumption and the facts and people it supposedly represents, the greater the risk that the policy could result in harm – and the harder it is to draw from the full spectrum of options necessary to build robust and equitable regulations.
None of this implies that policy assumptions are, in themselves, bad for AI policy discourse. But the most transformative and equitable AI policies will be those that engage actively with all uncomfortable counterpoints and with all under-represented perspectives – not just with the most prominent and powerful stakeholders.
To offset the potentially harmful effects of unchallenged assumptions about AI, stakeholders should:
- Recognize when an assumption that lacks firm, unequivocal evidence is being used as the basis for a policy, and provide a framework to consider that assumption’s consequences and counterpoint(s);
- Identify who these assumptions serve and consider whether those groups or individuals are representative of all stakeholders;
- Explore alternative or additional policymaking assumptions; and
- Establish a practice of pre-development risk assessments for proposed AI systems.

Next chapter

Topic themes

Defence and security

Economics and trade

Environment

Health

Institutions

Major powers

Politics and law

Society

Technology

Regions

Africa

Americas

Asia-Pacific

Europe

Middle East and North Africa

Russia and Eurasia

Foreign policy and international affairs events

Publications

Become a member

The Queen Elizabeth II Academy for Leadership and the Next Generation

About us

Recalibrating assumptions on AI

Arthur Holland Michel

Summary

US ownership of TikTok won’t protect democracies from digital threats

Where scientists can move the needle on global AI safety collaboration

The case for expanding digital public infrastructure

The EU’s new AI code of practice has its critics but will be valuable for global governance

Putting the Digital Services Tax on the table in US negotiations sends worrying signal on UK digital sovereignty

The internet under attack

In the face of growing AI cyber threats, do middle powers have agency?

AI export controls are not the best bargaining chip

How a surge in defence and dual-use technology investment could reconfigure the global AI race