Decision-Making for Impact: A Guide

https://www.lesswrong.com/posts/8KewJh3PfwKqcDugm/decision-making-for-impact-a-guide

Link post Contents

Begin with strategy

It’s important to place decisions in context. Decision theory can’t tell you what goals ought to be motivating you or who should be accountable for making decisions. So I like to tell people that decision-making doesn’t replace strategy. Instead, decision-making is the *manifestation *of strategy. Decision analysis methodologies are most powerful when you already have a pretty good idea of what you’re trying to do and how you want to try to do it. They work best when the challenge in front of you is to figure out how to take the general direction and principles you’ve already committed to and translate them into action in this specific situation. Accordingly, before you make use of the advice in the rest of this article, it’s best to have your mission, vision, and values figured out and committed to paper or pixels. You should know who has final authority for decision-making and, if decisions must be made as a group, how conflicts will be resolved among the group members. And you should ideally have some sense of the landscape in which you’re operating and the specific, near-term outcomes that are most critical to your success. Now, it might not be immediately clear why anything more than this is necessary. If you’ve already set out a general strategy, won’t its application to day-to-day decisions be straightforward? In some cases, it might be. Not every decision is a head-scratcher, after all. But almost inevitably, tradeoffs come up between goals or values that can’t be accommodated at the same time or to the same degree. Almost inevitably, people on your team or in your brain trust will disagree about the likely consequences of different paths forward, with no easy way to determine who’s likely to be right. And almost inevitably, there will be situations where the right choice will piss off all the wrong people. Decision analysis was made for moments like these.

Deciding what to decide

We make hundreds of decisions every day, from what to cook for dinner to whether to open up Twitter right now to how to respond to the slightly offensive comment your coworker just made. Most of these are hardly worth a moment’s thought, to say nothing of hiring a consultant to help you figure them out. If the discipline of decision analysis makes anything clear, though, it’s that some decisions matter more than others. Decisions in which the stakes are high, the context is unfamiliar, and the definition of success is not immediately obvious are the ones that will benefit most from your sustained attention. Due to the fast pace at which many teams operate, though, we often only start paying attention to decisions shortly before they need to be made. As a result, we shortchange many of our most important decisions and risk falling prey to cognitive biases that lead to worse outcomes. We can’t afford to thoroughly analyze every single decision we make, of course; we would quickly find ourselves so consumed by the exercise that we are unable to get anything done. So we need some way of anticipating and prioritizing our decisions in order to ensure we are spending the appropriate bandwidth on the ones that really matter. Fortunately, there is a simple activity called a decision inventory that can fix this problem. Adapted from a method created by decision analysis pioneers Strategic Decisions Group, the decision inventory is a means of cataloguing, consolidating, and sorting all of the major decisions you expect to make in the near to medium term. The exercise offers foresight on what dilemmas you will need to focus on when, and which ones to resolve quickly vs. investigate thoroughly.

Breaking down your decision

Okay, now we’re getting to the good stuff. Once it’s clear which decisions are most deserving of your attention, it’s time to start analyzing the ones on top of the pile. There are a number of ways one could potentially break down a decision into its component parts. My own process incorporates the following:

Going mental with decision modeling

The preliminary analysis described in the previous section will give you a holistic overview of your decision dilemma, but it optimizes for breadth over depth. It’s like exploring the world by looking at a two-dimensional map — you can see the complete picture, but not all the details. To achieve both, it’s necessary to model your decision explicitly using mathematical tools. This is like exploring the world from your desktop using a combination of Google Maps and Street View. The map helps you understand the broader context, but you can improve your understanding of the details by zooming in on specific points of interest. My invocation of technology here is intentional, because building decision models is a technological enhancement to human decision-making. We are leveraging the power and speed of modern computing capacity to help us gain a deeper understanding of our decision dilemma than would be possible using our minds alone. You are still in the driver’s seat — you are still the one who makes the decision, after all — but the mathematical model we create serves as a kind of supercharged coach or advisor. The nuts and bolts of putting together a decision model are too involved to cover here, but I wrote a handy primer featuring a grantmaking case study if you’re interested in learning more about how to do it. Sample decision model focusing on grantmaking for vaccine distribution (full article)There are two main reasons to put together a decision model. The first is that it’s possible to get much better precision about the factors at play and which ones are really important. When putting together a decision model, I use a statistical technique called Monte Carlo simulation that dynamically generates thousands of potential futures based on the estimates and assumptions that you provide (and can easily tweak in real time). Developed by nuclear physicists in the middle of the 20th century, Monte Carlo simulations have been empirically tested and shown to yield better judgments than alternative strategies in real-world settings. It’s the method that NASA swears by when preparing for so-called "complex projects" like launching human beings into space. The second awesome superpower of quantitative models is that they can help guide a research or learning agenda created for the express purpose of reducing uncertainty about important factors relating to the decision in question. You can use the concept of value of information to pick out the most important things to measure, measure them, and re-run the model in an iterative cycle until there is literally no more value to wring out of seeking further data. This method is called Applied Information Economics and was invented by information scientist Doug Hubbard for use in high-stakes investment decisions in the social sector and business world alike (read about a couple of examples of its use here and here). As far as decision analysis techniques go, AIE is pretty much the gold standard approach. With that said, modeling does have some costs and drawbacks. The main challenge is that, compared to the quick and easy process described in the previous section, modeling does usually require a more significant minimum investment of time and expense, an investment that scales with the number of stakeholders that need to provide input. In addition, modeling may feel somewhat intimidating or unnatural at first to staff or partners who have little experience with quantitative methods. Finally, while I’m a firm believer that anything worth modeling can be modeled, it is true that some kinds of goals and decision factors are harder to translate into mathematical terms than others and may require more specialized expertise. So when should you make the leap to modeling your decision? Ultimately, it comes down most of all to the stakes of the decision. Although all such situations should be judged case by case, a good-enough rule of thumb is that if your decision has more than $25,000 (for an individual) or $50,000 (for an organization) riding on it, it’s worth considering putting together at least a quick-and-dirty quantitative model to help ensure you get it right. And if the stakes are in the $250,000 and up range, I’d say it’s actually pretty irresponsible *not *to attempt a quantitative treatment of the situation. Finally, remember: the math is there to help you make your decision, not to make it for you. You are free to ignore it if you want. When I’m making a decision model I work hard to make sure it accounts for everything the decision-maker cares about. But I never deliver a recommendation without also considering the qualitative analysis that we developed alongside it.

After the math, the aftermath

Congratulations, you’ve made your decision! Give yourself a pat on the back and a well-deserved breather. But don’t take too long, for while the dilemma itself may be resolved, the consequences of your choice are just beginning to be felt. Here are some helpful questions to ask and things to think about in the post-decision phase:

You, too, can decide like a pro

So that’s how I help clients make decisions. If you have grappled with how to make smarter decisions for your organization or philanthropy and come up with similar or different techniques, I’d love to hear from you. There is always more to learn, and we are all in this together. Happy decision-making!