Fish Welfare Initiative recently concluded our latest re-evaluation point. The idea behind these “reval points” is to divide our work between planning and execution, such that we have times specifically designated for each and neither interferes with the other.
In this reval point, we decided upon a new 2-4 year strategy for our work in India. However, in the immediate term, this strategy change amounts to a more of a mindset shift than a programming/intervention shift: We are still collaborating with corporate and government partners, and we are still running our Alliance for Responsible Aquaculture.
What is changing though is why we are doing these things: Previously, we did them because we thought that they were the best programs to run in terms of having a short-term impact. Now, largely driven by our uncertainty (e.g.) about whether we had chosen the right programming, we primarily do them because we want to validate a scalable theory of change for helping fish in India.
On some level, this means that our mental energies right now are shifting more from scaling and helping fish in the short term to validating and helping fish in the long term. Though of course, we would always prefer to be scaling and helping more fish right now, we believe that this shift will best enable us to maximize our impact over the longer term.
The below post discusses our new strategy, why we chose it, and what this means for our work going forwards. Feedback is, as always, very welcome!
FWI’s Previous Strategy: Choose, Implement, Scale
Previously, FWI functioned more like how a typical organization would operate: We would choose goals and programming based on what seemed most promising, attempt to implement them, and then re-evaluate from there. As we generally set ambitious goals and usually attained them, this approach was in many ways very useful: For instance, in 2021 it enabled us to improve the welfare in 58 fish farms, secure India’s first corporate commitment for fish, and improve the lives of roughly 400,000 fish (see more). Our idea was that we broadly knew the right programming to do, and scaling was largely (though not completely) a matter of just applying more resources to replicate what we already had achieved at a smaller level.
However, we gradually began to think that this analysis was predicated on an overconfidence in our current work—a change was needed.
Motivations Behind the Shift
This shift in our strategy was motivated primarily by two reasons:
First, our leadership team has had the lingering feeling that our previous approach and goals were not on track to develop a scalable avenue for helping fish in India. Specifically, we realized that we were taking too many factors for granted, factors that need to function effectively if we are to scale and whose efficacy has been cast into greater question. In particular, Version 1 of our welfare standard is less impactful than we would like, and though we have long had the intuition that government and corporate avenues will enable our scaleup, we have thus far lacked a validated model in our context for how exactly this would happen.
And secondly, we have been thinking more explicitly about program development, particularly inspired by our consultant Karen Levy. Karen has previously consulted for GiveWell-recommended charities, charities whose programming and effectiveness inspire much of our vision. Her guidance has made us reflect more deliberately on what it will take to build a highly scalable and impactful program, and we felt that our previous strategy was inadequate to do so.
FWI’s Revised Strategy: Develop and Validate a Scalable Theory of Change
Our revised strategy is summed up by the following:
Focus less on choosing and implementing whole programs that seem promising.
Focus more on investigating the links in our best-guess theory of change and testing those, ideally by executing them on a smaller level.
Scale only after the theory of change has been well validated.
Be realistic about uncertainties where we have them.
Our current, high-level hypothesis of a scalable theory of change for helping fish in India is the following:
Note that this is broadly how we were thinking before, even though weren't explicitly focused on validating it. We were also seriously neglecting the farmer behavior aspect—now, as probably should have been obvious to us at the time, it is abundantly clear that thinking deliberately about what farmers are and are not willing to do, and what incentives are required at each step of the way, is critical for empowering them to change.
Zooming in to the lower level, in 2022 we specifically aim to investigate and validate each of these four links (two from institutions: one government, the other corporations), and then scale from there.
These changes have been reflected in our OKRs for 2022, which now include a more deliberate focus on testing each link. For instance:
Objective 2: Develop a subtheory of change of how welfare improvements influence fish welfare.
OKR 2.1: Welfare standard version 2 developed and published.
Objective 4: Develop a subtheory of change of how corporations influence farmers.
OKR 4.1: Trial completed of a corporation selling higher welfare fish. (This means that at least one corporation is selling fish we consider to be higher welfare, an occurrence which would validate our hypothesis that we can get corporations to change their purchasing and standards behavior.)
One potential downside of our new strategy is that, with its greater focus on preparing to scale later over immediately scaling now, we could be tempted to always push off scaling until some later tomorrow that may never arrive. Such infinite delays would be of course mistaken, as there are limits on how much uncertainty can be reduced—at some point you have learned what you can and the time for action arrives.
We are addressing this 'risk of infinite knowledge building' in the following two ways:
First, while we increase our focus on knowledge building we will also continue to help fish via our producer program, through which we already estimate that we have improved the lives of nearly 500K fish. We are at a stage now where we think we should almost always be having some impact, if only because there’s hardly a better way to learn about helping fish than by simply doing it.
Second, we here and now make the following public predictions about our timelines. We ask that our supporters hold us to account for these, including withdrawing their support if we do not stay true to our word:
We will strongly consider scaling down or pivoting our India operation, and publish a blog post explaining our reasoning, if either of the following occurs:
By January 1, 2024, we have not published Version 2 of our India welfare standard.
By January 1, 2025, we have not helped 500,000 fish with our Version 2 India standard.*
Thank you for helping us help fish!
* June 23, 2022 update: We lowered this latter lower bound tripwire from 1M to 500K. We did this because our current rate of helping fish is about 500K per year, and we think if we were still at that rate in 2024-2025 (assuming the worst case scenario of only having the V2 standard in January 2024) we would likely think that was still a project worth running.
* February 2, 2024 update: The first deadline has come and gone, so how did we do? In summary, a mixed bag: We did publish our Version 2 Standard at the beginning of 2023. However, over several in-field tests it did not seem to be better than our Version 1 Standard (V1), so we have instead stuck with V1, with which we estimate that we have helped over 500,000 fish since the original publication of this post. Furthermore, we have recently pivoted to begin looking into alternative program structures to the current ARA intervention altogether—learn more here. We consider this pivot, and with it the new leadership leading our R&D Department, to be in line with this original accountability metric.