What beats intuition when it comes to doing good? (with Marcus Davis)
Marcus A. Davis, co-founder and CEO of Rethink Priorities, discusses comparing diverse charitable outcomes, addressing moral uncertainty, and applying rigorous modeling to philanthropic decisions to maximize positive impact.
Deep Dive Analysis
13 Topic Outline
Comparing Different Charitable Outcomes
Cost-Effectiveness in Rich vs. Poor Countries
Arguments Against Comparing Charities
Plausibility of Moral Theories
Dealing with Moral Uncertainty
Aggregating Across Diverse Moral Views
Applying Explicit Models to Big Decisions
Common Pitfalls of Explicit Models
Balancing Simple vs. Complex Models
Modeling Personal Life Decisions
Motivation for Effective Altruism
Inefficiencies and Opportunities in Charity
Final Advice on Improving the World
11 Key Concepts
Commensurability of Outcomes
The ability to compare and evaluate different types of charitable or moral outcomes against each other, even if they appear vastly different (e.g., saving lives vs. funding art).
Diminishing Marginal Returns
The principle that as an individual or entity acquires more of a good (like money), the additional utility or benefit gained from each subsequent unit of that good tends to decrease.
Value Systems
The set of moral or ethical principles and beliefs that guide an individual's or group's judgments about what is good, right, or important.
Reflective Equilibrium
A state of coherence in moral philosophy where one's general moral principles and specific moral judgments are brought into alignment through a process of mutual adjustment and revision.
Moral Uncertainty
The state of not knowing which moral theory or set of values is correct, leading to difficulty in determining the 'right' course of action.
Aggregating Moral Views
The philosophical and practical challenge of combining different moral theories or perspectives, each with varying degrees of credence, to arrive at a unified decision or allocation of resources.
Empirical Uncertainty
Uncertainty related to factual outcomes and the real-world effects of interventions, such as whether a charity's program actually achieves its stated goals.
Normative Uncertainty
Uncertainty about what values truly matter and how much they matter, such as the relative importance of saving a life versus providing a positive experience.
Metanoronative Uncertainty
Uncertainty about how to make decisions when faced with both empirical and normative uncertainty, specifically how to aggregate different moral views or handle risk.
Explicit Models
Decision-making tools, often quantitative (like spreadsheets), that clearly lay out assumptions, inputs, and calculations, making the decision process transparent and auditable.
Implicit Models
Decision-making processes that rely on intuition, gut feelings, or non-transparent mental calculations, where the underlying assumptions and logic are not explicitly stated or examined.
14 Questions Answered
Yes, even if two charitable outcomes seem vastly different, pushing extreme examples often reveals that people can agree on which outcome is more valuable, making comparison possible.
Services and labor cost significantly less in low-income countries, and governments in richer countries often address problems like lead exposure, meaning charitable interventions can achieve greater impact per dollar in poorer regions.
This argument assumes morality is maximally demanding, which most moral theories do not claim; instead, they suggest dedicating a portion of resources to doing good, not eliminating all other valuable activities.
It's a complex process involving investigating logical arguments, premises, and implications, balancing argument strength with intuitive feelings, but ultimately, humility is key as definitive conclusions are hard to reach.
While the existence of objective moral truth is a contested philosophical question, the speaker suggests that given the weak evidence in philosophy, one should not be overly certain and instead consider theories that emphasize the importance of doing good.
One approach is to aggregate across different moral views by assigning credences to each theory and combining their recommendations, acknowledging that different aggregation methods can lead to different outcomes.
For individual donations, especially smaller amounts, it's often best to use rigorous charity evaluators like GiveWell, as they identify highly efficient and evidence-based interventions.
For big, complex decisions, explicit models are generally preferred because they make assumptions transparent and allow for systematic analysis, which is often more reliable than intuition in complex scenarios.
Be mindful of common failure points such as treating point estimates as certain, ignoring model uncertainty, and not questioning assumptions; instead, use distributions for uncertain inputs and continuously challenge the model's premises.
It's best to start with the simplest model that captures the core phenomenon and then gradually add complexity as needed, as simple models are easier to introspect on and identify issues.
While models can be useful for some personal decisions (like buying a car or renting an apartment), for highly subjective or dynamic personal choices (like having a child), they may not be as applicable or straightforward.
The motivation stems from a belief that it's possible to make the world better, a sense of owing it to past generations who improved the world, and the personal satisfaction of working on impactful problems with smart, caring people.
While it's harder than pursuing immediate personal happiness and involves navigating empirical and philosophical uncertainties, it's not incomprehensibly complex, and significant progress can be made through rigorous analysis.
Yes, due to fewer financial incentives in the charitable sector, there's less competition, allowing individuals and organizations who apply rigor and analysis to make a disproportionately large positive difference.
10 Actionable Insights
1. Prioritize Location for Impact
When addressing a specific problem like lead poisoning, focus on low-income countries where interventions are significantly cheaper and more cost-effective, yielding greater impact per dollar.
2. Don’t Avoid Comparing Good
Reject the notion that radically different forms of good (e.g., art performance vs. saving lives) are incomparable; pushing extreme examples often reveals that most people agree some outcomes are clearly more valuable.
3. Be Humble About Moral Theories
Recognize that philosophical evidence for specific moral theories is weak and contested, so avoid being overly certain about one particular view.
4. Aggregate Across Moral Views
For large-scale giving (e.g., foundations), combine different moral theories (e.g., utilitarianism, deontology, virtue ethics) using weighted averages or proportional splitting to account for moral uncertainty.
5. Use Explicit Models for Big Decisions
For complex, high-stakes decisions, build explicit models (like spreadsheets) to make assumptions transparent and track interactions, rather than relying solely on vague intuition.
6. Model Uncertainty, Not Point Estimates
In explicit models, represent uncertain inputs as distributions (e.g., using Monte Carlo simulations) rather than single point estimates to avoid underestimating the true range of possible outcomes.
7. Start Simple, Then Refine Models
Begin with the simplest model that captures the core phenomenon, then gradually add complexity and additional factors as the stakes increase or as you identify critical areas for deeper analysis.
8. Integrate Judgment with Models
Do not blindly follow model outputs; instead, use models to clarify thinking, identify disagreements, and inform judgment, always considering what might be missing or where assumptions could be flawed.
9. Leverage Charity Evaluators for Donations
For individual donations, especially smaller amounts, use highly rigorous charity evaluators like GiveWell to identify evidence-based, efficient charities that maximize impact.
10. Differentiate Personal vs. Altruistic Decisions
Recognize that the moral constraints and decision-making approaches for personal life choices (e.g., hobbies, family) may differ from those for large-scale altruistic giving.
6 Key Quotes
not all value systems are equally plausible.
Marcus A. Davis
Once you admit there's some case it doesn't work, then we're arguing about like, why is that the case?
Marcus A. Davis
just because someone can like say, yeah, that's just my belief that, uh, these things are incomparable. That doesn't mean like I have to like take that seriously.
Marcus A. Davis
I don't think you should be that certain one way or another about this question.
Marcus A. Davis
all models are wrong, some are useful.
Marcus A. Davis
Progress is possible. Even when hard questions in charity, in thinking about how to make the world better, you can do better than your intuition.
Marcus A. Davis