1 of 1

v

v

v

We need to improve how we develop and use AI to ensure long-term human flourishing.

Behaviour science can help.

Alexander K Saeri�alexander.saeri@monash.edu

Four behavioural lenses on AI as a socio-technical transition

Our paper, Behaviour in sustainability transitions, described how behaviour science can understand and accelerate changes in systems for better futures.

The Centre for the Governance of AI advances a research and action agenda for navigating the transition to a world with advanced AI.

Holden Karnofsky (Open Philanthropy) argues that this century is especially important because AI could dramatically speed up scientific and technological advancement, so we should figure out how to make sure that goes well.

Beyond contemporary policy challenges, the long-term risks & opportunities of AI are enormous.

A ‘general-purpose’ or superintelligent AI would be transformative, eclipsing steam power / electricity.

1

The development, spread, and use of AI involves both technological and social change: a ‘socio-technical transition’.

These changes are interdependent and will influence each other.

Societies and policy makers struggle to respond and manage these transitions because they are complex, fast paced, and involve many different system actors, incentives, and behaviours.

If we can understand and change decision-making & behaviour, we can develop and use AI to maximise the chance of long-term human flourishing.

Behavioural lens

Example in AI

Who needs to do what differently?

Citizens advocate for legislation & common standards for beneficial use of AI

What are the biases, habits, and moments that matter?

Recommender & ‘expert’ AI systems reproducing societal bias and fostering polarization

Who wins and loses?

Corporate and state ‘race to the bottom’

What are the emerging innovations and social practices?

AI organisation governance and disclosure norms

AI governance & behaviour science: find out more

How can we best navigate the transition to a world with advanced AI?

Time

Human well-being & flourishing

Good human decision-making and behaviour about how to develop & use AI

Poor human decision-making and behaviour about how to develop & use AI

We are here

2

3

4

doi.org/10.1016/�j.eist.2021.10.010

governance.ai/about

cold-takes.com/most-important-century/