Negotiations and pressure campaigns have proven effective at driving corporate change across industries and movements. I expect that AI safety/governance can learn from this!

The basic idea:

• The runaway success of effective animal advocacy has been sweeping corporate reform
• Similar tactics have been successful across social movements
• GovAI to ??? to PauseAI: Corporate campaigns need an ecosystem of roles and tactics

Possible next steps

• Pragmatic research: ask prioritisation, user interviews, and message testing
• Start learning by doing
• Work extensively with volunteers (and treat them like staff members)
• Moral trade: longtermist money for experienced campaigner secondments

The runaway success of effective animal advocacy has been sweeping corporate reform

Animal advocates, funded especially by Open Philanthropy and other EA sources, have achieved startling success in driving corporate change over the past ~decade. As Lewis Bollard, Senior Program Officer at Open Philanthropy, writes:

A decade ago, most of the world’s largest food corporations lacked even a basic farm animal welfare policy. Today, they almost all have one. That’s thanks to advocates, who won about 3,000 new corporate policies in the last ten years…

In 2015-18, as advocates secured cage-free pledges from almost all of the largest American and European retailers, fast food chains, and foodservice companies. Advocates then extended this work globally, securing major pledges from Brazil to Thailand. Most recently, advocates won the first global cage-free pledges from 150 multinationals, including the world’s largest hotel chains and food manufacturers.

A major question was whether these companies would follow through on their pledges. So far, almost 1,000 companies have — that’s 88% of the companies that promised to go cage-free by the end of last year. Another 75% of the world’s largest food companies are now publicly reporting on their progress in going cage-free.

Some advocates establish professional relationships with companies and encourage them to introduce improvements. Others use petitions, protests, and PR pressure to push companies over the line.

Almost everyone who investigates these campaigns thoroughly seems to conclude that they’re exceptionally cost-effective at making real improvements for animals, at least in the short term. There are both ethical and strategic reasons why some animal advocates doubt that these kinds of incremental welfare tactics are a good idea, but I lean towards thinking that the indirect effects are neutral to positive, while the direct effects are robustly good. There are other promising tactics that animal advocates can use, but the track record and evidence base for corporate welfare campaigns is unusually strong.

Of course, animal advocacy is different to AI Safety. But something that has been so successful in one context seems worth exploring seriously in others. And oh wait, it has worked in more than one context already… 
 

Similar tactics have been successful across social movements

In my research into other social movements’ histories, I found strong evidence that pressure tactics can be effective at changing companies’ behaviour or disrupting their processes^[1]:

• US anti-abortion activists seem to have successfully disrupted the supply of abortion services and may have reduced abortion rates.
• Anti-death penalty activists successfully disrupted the supply of lethal injection drugs.
• Pressure campaigns likely accelerated Starbucks and other chains’ participation in Fair Trade certification schemes.
• Prison riots and strikes seem to have encouraged the creation of new procedures and rules for prisoners.

There are lots of caveats, concerns, and intuition-building anecdotes in each of the case studies. But I wanted to highlight the general pattern.
 

GovAI to ??? to PauseAI: Corporate campaigns need an ecosystem of roles and tactics

Successful corporate reform campaigns in animal welfare and other movements suggest an important principle — the need for an “ecology” of different organisations, roles, and tactics.

In animal welfare campaigns, ‘good cop’ orgs who work closely with industry (e.g. Compassion in World Farming) are complemented by ‘bad cop’ orgs willing to use pressure tactics (e.g. The Humane League).

There’s scope for a broader spectrum still, from professional, expert groups for credibility (e.g. animal welfare scientists, vets, nutritionists) through to grassroots activist and mass mobilisation groups (e.g. Direct Action Everywhere, Animal Rising). It seems like AI governance is doing better at these extremes. Most of its weight is behind the ‘credible experts’ with groups like GovAI, CSER, and CSET — which I do think is the best place to start. 

Some orgs have started to spring up at the other end of the spectrum too, like the Campaign for AI Safety and PauseAI^[2]. This seems good to me! Having organisations that use radical tactics seems to increase identification with and support for the more moderate groups. Many academics with expertise relevant to social movements seem to think that “having both radical and moderate flanks” and “the strategic use of nonviolent disruptive tactics” are quite important for social movement success. 

But we’re missing the space in between these extremes that focuses on consistent pressure on companies, roughly what the Anyi Institute calls “structure organizing”; working “to build an organized base of people to pressure decision makers around certain demands.”^[3]
 

The model that seems to be missing is at the overlap of three things^[4]:

• Structure: a nonprofit with permanent, full-time staff applying consistent pressure on particular companies (not flash-in-the-pan or solely volunteer-led)
• Audience: targeting companies (not consumers, or governments). You might do some public outreach to recruit more activists or raise the stakes for the companies you target, but the goal is to feed the specific campaign.
• Some confrontation: you can start with the carrot (good for the world, good PR), but you might need to resort to the stick (bad PR, wasted company time, some missed profit) if they resist. Similarly, you can start with friendly meetings; you might need to get public.

There are lots of groups in effective animal advocacy that sit in this space. To name a few^[5]…

• The Humane League
• Mercy For Animals
• Sinergia Animal
• Anima International
• Animal Equality
• Essere Animali
• Vegetarianos Hoy
• Equalia
• Humánny pokrok

Caveats and concerns

There are many valid caveats and concerns to the general argument above, but I don’t think any of these are sufficient to overturn the conclusion.

Here are some brief ideas:

• We have uncertainties about proposed governance asks… but some seem promising
• This seems confrontational… but nonviolent confrontation has its place (also), and you don’t always even need to resort to confrontation^[6]
• This is more technical than most pressure campaigns… but safety (plus equity and employment) concerns have underpinned past successful campaigns on tech issues^[7]

Let me know in the comments if you have other concerns or would find it helpful if I fleshed out my counterarguments to any of the above!

Possible next steps

This isn’t meant to be anything like a full “roadmap”, but I wanted to provide at least a smattering of suggestions for next steps.

Pragmatic research: ask prioritisation and message testing

Research from groups like GovAI is mostly academic or advisory in style. Of course there’s an important role for that. But to prioritise actions, you need something more pragmatic.

I’m a big fan of Charity Entrepreneurship’s process for generating lots of possible intervention ideas, then narrowing the scope down to identify those that seem most promising. Rethink Priorities’ Existential Security Team does something similar.

An alternative way to contribute to pragmatic research is through message testing. There’s been some work on this already, like by Vael Gates and Campaign for AI Safety, but there’s room to get more systematic (compare Pax Fauna on animal advocacy messaging) and rigorous (compare Schubert, Caviola, and Faber on the psychology of x-risk).
 

Start learning by doing

Desk-based research isn’t enough. You need to get out there; start meeting your stakeholders and encountering real conditions, needs, bottlenecks.

Fish Welfare Initiative’s top “lesson learned” from their first 2.5 years was that “It’s often better to explore by exploiting (as opposed to theoretical desk research)”:

Initially we had a fairly discrete research process, where we aimed to answer the following 4 big prioritization questions separately via desk research: (1) Which (fish) species should we work to help? (2) Which country should we work in? (3) Which welfare improvement(s) should we make? (4) Which approach(es) should we take?... However, the reality is that, however hard we tried to answer these questions separately… these questions are all interlinked—you can’t, for instance, work on Atlantic salmon welfare in India because there are no Atlantic salmon farmed here. This is something we would have better internalized had we gone out into the field earlier (for instance to actually visit farms and talk with companies in the countries like India where we were interested in working)... we spent about a year trying to answer our prioritization questions separately, and largely doing it through desk research… we believe we could have cut that down to roughly half the time. Now, one of our organizational mottos is “explore by exploiting”: We often think the best way to learn how to do something is just to try to do it. The true bottlenecks usually become clear in the attempt.

This maps on pretty well to the absurdly consistent startup advice to do regular user interviews. Although the style of conversation looks different, you can use these to understand:

• Value (needs and demand)
• Usability (e.g. whether a certain format or message makes sense)
• Feasibility (e.g. whether you can actually offer what’s needed)

But even this can only take you so far; you also need to actually prototype or pilot “minimum viable products”. This is probably easier with tech products than with complex social issues, but the underlying sentiment has merit.
 

Work extensively with volunteers (and treat them like staff members?)

I have the sense that a bunch of longtermists have internalised the opening message of this old 80k post a bit too hard and think that volunteering is ~useless.

But try telling that to effective animal advocacy groups; I suspect many of them would collapse without this kind of unpaid support. Corporate campaigns benefit from both professionals and volunteers.

The more obvious form of volunteering is the quick or low-effort kind, like online activism or joining ad hoc protests. But consider also treating some volunteers like staff members, with defined responsibilities, benefits, and managers.
 

Moral trade: longtermist money for experienced campaigner secondments

Crudely, AI safety seems bottlenecked more by expertise than by money, while animal advocacy has expertise but lacks money. Some sort of swap (moral trade) seems possible.

Of course, this is a simplification. But AI Safety has tended to be bottlenecked more by various types of talent (plus management capacity, limits on the speed of growth, etc) than by funding.^[8] It’s a mixed picture in animal advocacy; talent is still needed, but funding seems a comparably big constraint.

I can imagine AI safety orgs doing something like paying for experienced animal advocacy campaigners to go on secondments or provide advice. Salaries are usually constrained by a bunch of organisational considerations that perhaps grants to other orgs are not — so maybe the orgs could pay what they value the support at (which could be very high figures, by animal advocacy’s standards)?

Of course, I can imagine a whole bunch of reasons why this might not work out in practice. Maybe I’m overcomplicating it and there are simpler answers:

• Put out a job ad with a decent salary. After all, corporate campaigns exist outside of animal advocacy, so you might get applications from experienced activists.
• Use existing resources and public advice by advocates or research orgs.
• Find informal advisors with experience from other movements, perhaps including effective animal advocacy.

Thanks to Patrick Levermore and Holly Morgan for feedback on drafts of this post.

1. ^{^}

   For sources and detailed evidence, see row 7 of this spreadsheet, which describes the relevant sections of each report.

2. ^{^}

    These orgs are new and still pretty tiny though. E.g. a recent CAS update notes that a UK petition only has 38 signatures. 38!? For reference, the two petitions I found via The Humane League UK’s site had 374,791 and 19,826 signatures, and it needs 10,000 signatures to get a government response.

3. ^{^}

    We could think about (partly overlapping) spectrums or portfolios of asks/demands, tactics, target audiences, and structures. I suspect a healthy social movement needs at least some groups at different positions on all of these.

4. ^{^}

    I think I conflated these three somewhat in this post; if I’d spent longer, ideally I would have argued for each of them separately. Let me know if you’d be keen for me to write a follow-up that does so.

5. ^{^}

    I made this list from memory, so apologies to any orgs I mischaracterised. Feel free to contact me if you disagree. Also, there are others that share the structure and audience but eschew the confrontational side, most notably Compassion In World Farming. And others that focus on replacing animal products with alternatives rather than animal welfare.

6. ^{^}

    If I recall correctly, when The Humane League UK first got set up, they won something like ~50 corporate commitments in their first year with only one staff member, mostly just by asking what no one had asked before. But their annual reports don’t start until their third year of existence, so I can’t quickly verify this.

7. ^{^}

    There’s also precedent for various different types of stakeholders, both internal and external, influencing AI companies.

8. ^{^}

    I don’t know of a good survey on this. There’s this but it’s a few years old and cross-EA. The situation has changed a lot since FTX’s collapse. And I’m not sure how much weight to place on this.