AI Literacy and Governance
29th December 2024, by Ismael Kherroubi Garcia
On 26th December 2024, Kairoi was featured in AI, Government and the Future, a podcast by Corner Alliance that dives into the latest AI advancements, government policies, and strategies helping shape the future of our society.
Watch the discussion on YouTube, or listen in on Spotify or Apple Podcasts.
The key topic of the discussion was the need for AI literacy amongst lawmakers. With more or less well-known gaffes at the intersection of technology and law —from Meta CEO being asked about sending emails via Whatsapp by a senator in 2018,¹ to lawyers citing fake cases from AI chatbots more recently² ³—, it is clear that there is a need for lawmakers to have a deeper understanding of the technologies underpinning AI, the human labour sustaining AI, and the societal impacts of AI tools.
The Message
Kairoi delivered one key but provocative message: the desire to make “black box” AI models “transparent” misses the role legislators have in holding individuals and organisations accountable.
It’s important to note that AI tools are opaque by design: they rely on complex internal computations that we can’t inspect. We can explain how such a tool works, but we can’t explain how individual outputs are produced. From a technical perspective, this is simply not generally possible. But this is not the crux of the problem. Focusing on the tool’s internal workings is not that valuable. Rather, consider the role of people throughout an AI tool’s lifecycle. Here, a first analogy can help.
Consider that legislators have a say in how we drive. Crossing a red light, not signalling when changing lanes, and driving over the speed limit are some ways to be fined as a driver (in the best case scenario). Meanwhile, the internal workings of the vehicle —how gas makes it into a chamber, how a piston moves, how gears change, what source of power is being used and why, and so on— need not be known by the police officer stopping the perpetrator, nor by the judge if the case is taken to court. Rather, what is legislated —in basic terms— is how individuals and organisations are held accountable. In the case of a car crash, an insurance company may have to step in. Through their investigation, it might be found that there was a factory fault with the vehicle, and the manufacturer may have to pay for it. With no need for an advanced engineering degree, all the above —insurance companies, consumer rights and traffic safety— are heavily regulated.
The analogy is not perfect, but the key message is simple: lawmakers should focus on holding people and organisations accountable for the errors of the technological artefacts we manufacture. With this, AI literacy as it pertains to regulating AI requires a deeper understanding of the human decisions that result in AI use cases —from the motivations of those funnelling investments into AI, to the full AI development lifecycle, and to its final use.
For more insights on AI governance and literacy, listen to the podcast linked above, and reach out to us to learn how Kairoi can support your responsible AI journey.
Contact us
hello@kairoi.uk
References
¹ Stewart, E. (2018) Lawmakers seem confused about what Facebook does — and how to fix it, Vox, online [accessed 29 December 2024]
² Brodkin, J. (2023) Lawyers have real bad day in court after citing fake cases made up by ChatGPT, Ars Technica, online [accessed 29 December 2024]
³ Steacy, L. (2024) A B.C. lawyer submitted ‘fictitious’ cases generated by ChatGPT to the court. Now, she has to pay for that mistake, CTV News Vancouver, online [accessed 29 December 2024]
