For the climate, little things don't add up

We need to think seriously about collective action

Suppose I said “I’ve found a cheat code to stop climate change. See, climate is a collective action problem. That means that we can’t just look at my personal emissions. We need to ask what would happen ‘if everyone did it.’ That’s why I’ve started pausing my microwave one second early. That might not save much energy, but if everyone in the world did it, it would save way more energy than I’ll use in my whole life! Little things really add up when you’re thinking about the climate, so I’m doing my part by pausing my microwave. I’m part of something that will stop so much emissions. It’s cool that it’s so easy to have such a giant impact.”

Something seems wrong here.

I think this is actually just another way to phrase the statement “Tiny things we do may not seem like much, but if everyone does them, they have a big impact on the climate.” Both seem equally silly to me, but the second is very common in conversations about climate change.

Tiny parts of our personal emissions are tiny parts of global emissions

I’m going to make an obvious point that I worry is glanced over in these conversations.

Suppose this is a single person’s emissions. The red represents the emissions of a tiny activity they do.

Some might say that this small red square will add up if a lot of people do it. What would several people doing this look like?

We can combine them together into a single clump of total emissions.

Gathering the red dots together:

Notice that the ratio of the red activity to all the other emissions is the same for the group as the individual person.

This obviously makes simple mathematical sense. 1/100 = 4/400. We’re just multiplying numerators and denominators by the same constant. The ratios themselves stay the same.

But I worry that this basic intuition is lost on a lot of people when they say things like “tiny things add up.” Tiny things add up absolutely, but they don’t add up relatively. They often remain the same tiny percentage of total emissions as they are of individual emissions. An activity that’s 1/10,000th of your personal emissions is likely to also be 1/10,000th of global emissions. It’s only as promising a way to reduce global emissions as it is to reduce your personal emissions. If you’re trying to rally everyone to stop an activity that wouldn’t meaningfully raise your personal emissions at all, this is a sign that you’re wasting your and others’ time attention and effort on a drastically ineffective thing to do in the name of stopping climate change.

AI fits this pattern perfectly. All data centers emitted 180 Mt CO2 in 2024. AI likely used about 15% of data center electricity, so it emitted around 30 Mt. Most of the emissions from AI come from electricity used in data centers, not from physical construction of AI hardware, so its total emissions weren’t much more than 30 Mt.

Globally, the world emitted 41,600 Mt CO2 in 2024. AI was responsible for 0.07% of emissions last year. This includes every single instance of AI. Not just chatbots, but literally every instance of deep learning (there are a lot of these). Prompting chatbots 100 times during the day would add about 0.007% to your daily emissions. The reason global AI is 10x higher as a percentage is a combination of the fact that you interact with many other AI applications throughout the day without realizing it, Americans emit way more than average, and the average global citizen isn’t interacting with AI (or other things) as much. You’ll notice that when you see AI compared to total emissions, its proportion of total emissions looks very similar whether you’re looking at an individual person using chatbots a lot or the world as a whole.

This shouldn’t surprise us, because tiny things on an individual level don’t add up to huge things on a global level, they stay the same tiny proportion globally as they are individually. The world is really only individuals putting different demands on the broader energy system, so this makes sense. A behavior change is only as promising a solution for climate at the global level as it is for your personal carbon footprint. Tiny things have equally tiny potential for actually moving the needle on how much we emit.

Making the list

What if you could make a list of demands for every behavior change you’d want everyone to make? The list would be ordered by which actions would help the climate the most if everyone did them. You’d want everyone to know about the first few and act on them. Maybe you could get the first 5 across to a wide audience. After that it starts to get more uncertain.

What would be at the top of the list? What would you prioritize? For me, we’d start with these three:

1. Find any ways you can push for converting the grid to green energy (I’d recommend Clean Air Task Force to get up to speed on where you can be most useful).

2. Donate to well-researched high impact climate charities.

3. Vote for politicians who will take significant positive actions on climate.

There’s a big power gap between these and any specific lifestyle changes you could make. Lifestyle changes are nothing in comparison to big systematic changes to our energy grids. The differences are not small. They’re gigantic. Here’s an example I used before to illustrate the point:

Suppose there are 3 people who each want to have an impact on the climate: Kate, Bob, and Freddy. They each independently choose their own ways of impacting the climate. All seem really noble and self-sacrificial. Kate joins a committee of 500 people working for a year to keep a nuclear power plant open for another 10 years. Any impact they have will be divided by 500 people. Bob goes vegan for a year. Freddy has a debilitating ChatGPT addiction and prompts it 2,000 times per day. About one prompt every 20 seconds every waking hour. He quits for a year.

Try to form an idea in your head of how their climate impact compares to each other. It’s not immediately obvious.

After 1 year, Kate has prevented 70,000 tonnes of CO2 from being emitted. Bob has prevented 0.4 tonnes. Freddy prevented 0.2.

This means that Kate had as much effect as 175,000 Bobs, or 350,000 Freddys.

Here’s how many people would need to go vegan for a year to match Kate’s impact:

I worry that when people sneer at quantifying climate interventions, they don’t realize how gigantic the differences are in what we can do.

To use a very blunt analogy, if we’re in a war against climate change, individual contributions to changes to the energy grid are like nuclear bombs, and individual lifestyle changes are like sticks of dynamite. It often seems ridiculous to me to even mention them in the same breath.

But let’s say you want to get those most important lifestyle changes across to people anyway. What would be most important? I’d push for relatively simple changes that aren’t especially disruptive to people’s lives but still cut huge amounts of carbon; the things most people can stick to that will actually make some kind of noticeable dent.

4. Buying green energy from your grid or installing a private solar panel.

5. Buying an electric car.

6. Skipping a flight if there’s another way of traveling.

7. Doing a deep one-time retrofit of your home.

Try making your own list like this. Think about what 20 interventions for the climate you would most want to communicate to the average person. Try to think about how many the average person could actually be convinced to follow. Where would you get the most climate impact for your effort?

Where would AI fit into this?

These are things the average American does every day that emit more than prompting ChatGPT 100 times (28 g CO2, and yes, including training and the embodied emissions of hardware):

• Making coffee

• Running an AC for 10 minutes

• Running a fridge normally

• Taking a 1 minute hot shower

• Using hot water for 1 minute to rinse dishes

• Ironing clothes

• Microwaving food

• Cooking on an electric stovetop for 5 minutes

• Leaving 8 LED bulbs on for an hour

• Running a humidifier for a few hours

All of these things would be way, way, way down my list of the most important things for Americans to cut for the climate. We have a limited number of ways the average person will be willing to adjust their lives. A limited amount of messaging we can get to them before they focus on other things. Only if I knew that I could literally get a majority of people to change their lives in hundreds of other ways for the climate would I then consider suggesting “being mindful of your AI prompts.” As of right now, the vast majority of Americans wouldn’t consider paying more than an additional $40/month in taxes to combat climate change. They are not ready to make hundreds of additional lifestyle changes, and if we get to a future of abundant green energy, we won’t need to ask for these changes anyway.

Worse, if an activity is hundreds of lines down this list, unless you’re sure that the person has also cut every single thing before it, it is very likely that if they cut the activity, they’ll replace it with something higher up on the list that emits more and is less useful to them. So going too far down this list puts you in more and more danger that you recommendation increases rather than decreases their emissions. Not only is it distracting and ineffectual, it might even raise the average person’s emissions.

Not many people are actually thinking about climate change much

One of the reasons it’s so important to get climate messaging right is that for a majority of people who hear it, it will be one of the only times they think about climate change at all. Only 23% of Americans say they have thought about climate change a lot.

Source

65% say they rarely or never discuss climate change with friends

Source

A majority of Americans report that their friends and family don’t believe it’s especially important to put effort into stopping climate change.

Source

And 44% don’t believe the actions of individuals will have any effect on the climate.

Source

Unfortunately, this does make it seem like climate messaging is somewhat zero sum: one message people hear may crowd out others. Each message will often be one of the only times the listener considers climate change. This means we need to communicate especially important information when we do recommend lifestyle changes. Sacrificing this rare important opportunity on a message that will cut literally millions of times fewer emissions than other possible lifestyle changes seems like a decision to crowd a very limited field with something that won’t help at all. Advertising extremely tiny cuts as significant is effectively a form of kneecapping real climate action. I see people going around warning about the impacts of personal AI use as irresponsible, in the same way I think it’s irresponsible to imply that you can have significant climate impact if you randomly unplug your fridge every now and then. Messages like this are crowding out serious climate conversations for the sake of something that will never be a significant part of our emissions. The competition to reach everyday people is already extremely crowded, and people waste valuable opportunities when they throw out innumerate recommendations that won’t save us from the worst effects of climate even if all 8 billion people alive ruthlessly followed them.

Collective action

Climate is a collective action problem.

We all have strong incentives to use more fossil fuels. Fossil fuels are, unfortunately, miraculous. It’s a miracle there’s a liquid you can spend a few minutes filling your thousand pound metal vehicle with that will singlehandedly blast it hundreds of miles. There’s no reason these had to exist, but they do, and because they’re so miraculous it’s very hard to make green replacements for them, so those replacements cost more money. Life’s easier if you personally don’t worry about climate change and just do what you want.

Your individual emissions will, on their own, literally never matter for the climate. You could fly in a private jet every day of your life, and your emissions wouldn’t make a dent in the amount that’s actually going to cause tipping points for the climate. No one in a hundred years will notice any difference in the weather based on your entire lifetime emissions.

However, if a lot of people don’t change their emissions, climate change will make life significantly worse. Larger and larger catastrophes will happen.

This is a classic collective action problem: Everyone would be much better off if we all took some action, but everyone individually has a strong incentive not to, and each person’s contribution won’t matter unless a lot of people do it. Thus, the only way we can really encourage people to cut their emissions is if we reliably show that many other people are doing the same, so there will in fact be collective payoff for individual sacrifices.

Because climate is a collective action problem, we should keep a few things in mind:

• People have strong reasons to emit. Asking them to change their behavior and make sacrifices for the climate is already a lot. We can’t presume they’ll have infinite willingness to change. If they did, we would have solved climate change already.

• We need to find simple rules that everyone can know everyone else is trying to follow. If they’re too complex, citizens will lose hope that others are doing their part of the collective action problem, and they themselves will be less likely to play along. These simple rules also need to be things that will actually reliably reduce emissions a lot. The rule “pepper your daily life with random impulses of guilt about some tiny emissions but not others. Cut based on ominous news stories rather than actual numbers” seems like a terrible simple rule to give people as a way of solving the collective action problem of climate, yet that seems to be a very popular message in a lot of spaces. We should discourage this.

For more on collective action problems I’d review the prisoner’s dilemma. We need to remember in climate communication that we are not merely building big cool coalitions of the virtuous in-the-know people, we are asking everyone to take scary leaps into potentially more difficult lives to solve a problem that will only really be solved if many other people take the same leap. This is hard! It’s why solving climate change is so difficult. We need to think about this as threading a needle. People who run around hyping people up about tiny emissions merely because they “add up” to still very tiny proportions of global emissions are not taking this problem seriously.

If we want to limit warming to 2 degrees by 2100, this is a bad time to do that.