PAUSE for AI

Where it began

PAUSE for AI grew from a project I developed as part of my work toward UCLA's Leaders in Sustainability Certificate while completing my Executive MBA at UCLA Anderson.

The starting question was deliberately small: could one simple change in how an AI assistant responds reduce unnecessary output without making the response less useful?

An early signal

The first exploratory pilot involved eleven volunteers using ChatGPT, Claude and Gemini. A separate standardised before-and-after benchmark was then run on Claude and ChatGPT using a one-time instruction designed to remove filler and limit unnecessary response length.

This was an encouraging signal, not a conclusion. It was not an intervention completed by all eleven volunteers, and it does not establish an equivalent reduction in electricity, water use or emissions. Those relationships depend on the model, hardware, data centre and energy system behind each interaction.

The question stayed with me after the certificate project ended. I am now expanding the work through a larger validation study across more participants, platforms, tasks and levels of AI experience.

See the study and its limitations →

Why I am continuing

Social media offers a useful warning. Its adoption moved faster than public understanding of how constant feeds and engagement-driven design might affect attention, behaviour and wellbeing. Early concerns were easy to overlook. By the time the conversation became mainstream, many of the habits and business models were already deeply established.

AI raises different questions, and the comparison should not be stretched too far. Its possible effects on cognition deserve a separate discussion. The immediate concern behind this work is also physical: the electricity and water required by systems whose use may grow much faster than public awareness of what sits behind each interaction.

I am not suggesting that individual users alone can solve AI's environmental impact. AI companies, infrastructure providers and policymakers carry much larger responsibilities. I am also not arguing for less useful AI. I use these tools and see their value.

The purpose of bringing users into the conversation is not to blame them or make them feel guilty. It is to give people enough visibility and understanding to make informed choices before patterns of use become invisible habits. Better infrastructure, platform accountability and thoughtful user behaviour should be parts of the same response.

The wider work

The experiment opened up three connected areas that PAUSE for AI is continuing to explore.

Reduce what is unnecessary

Can better defaults and more intentional use reduce avoidable AI output while preserving usefulness?

Make resource use visible

Could AI platforms give people a practical view of the resources associated with their usage, much like an electricity bill or a weekly screen-time report makes an otherwise invisible pattern easier to see?

Keep the human in the thinking

How can AI literacy help people use these tools to extend their thinking without gradually outsourcing the judgement, curiosity and struggle through which thinking develops?

An invitation, not a finished answer

This work is still developing. I am sharing it openly because useful ideas become stronger when knowledgeable people question their assumptions, test their methods and point out what they may be missing.

If you work in sustainability, behavioural research, AI, education or public policy, I would value your critique. I would also be glad to hear from institutions or communities interested in participating in the next stage.

Share a critique or explore collaboration →

Use intelligence wisely, both the machine's and our own.

PAUSE for AI is an independent initiative that grew from a UCLA Leaders in Sustainability certificate project. It is not presented as an official UCLA research study or UCLA-endorsed programme.

← Back to Ideas in Practice