Effective Altruism: Critique and Value
Also known as:
Effective altruism uses evidence to maximize impact; critics note it can devalue relational giving and overlook systemic change. Holding both instrumental effectiveness and relational values is possible.
Effective altruism uses evidence to maximize impact, yet holding both instrumental effectiveness and relational values within the same commons is possible.
[!NOTE] Confidence Rating: ★★★ (Established) This pattern draws on Moral Philosophy.
Section 1: Context
The commons for giving and social change is fragmenting along measurement lines. On one shore: funders, researchers, and mission-driven organizations deploying rigorous metrics—cost per life saved, disability-adjusted life years, counterfactual impact. On the other: communities, mutual aid networks, and long-relational practices built on belonging, reciprocity, and deep ecological knowing that resist quantification. Neither ecosystem is stable. The measurement-first approach generates funding velocity but accumulates legitimacy debt with the communities it aims to serve. The relational approach sustains deeper trust but struggles to coordinate across scale or attract institutional resources. In corporate contexts, this manifests as impact investing frameworks that miss cultural value. In government, it appears as evidence-based policy divorced from constituent voice. In activist spaces, it’s the tension between data-driven campaign targeting and ground-rooted relationship-building. Tech platforms now mediate both: algorithms optimize for measurable outcomes while erasing the relational substrate that made giving meaningful. The pattern arises because both approaches contain real truths—measurement does prevent waste, and relationship does prevent harm—yet the system treats them as zero-sum rather than co-generating.
Section 2: Problem
The core conflict is Effective vs. Value.
The effective altruist impulse asks: How do we know we’re not wasting resources? How do we scale good to the largest number? It demands evidence, attribution, and fungible metrics. The value-centered impulse asks: Who decides what counts as good? How do we preserve the agency and dignity of those we serve? What becomes possible when we give from relationship rather than obligation?
When effectiveness dominates alone, the commons develops brittle architecture: relationships become instrumentalized (you matter because your problem is measurable), communities experience extractive research (data harvesting without reciprocal learning), and giving becomes depersonalized (a calculation, not an act of presence). The system gains velocity but loses root. When relational values dominate alone, the commons cannot scale beyond trusted networks, cannot hold itself accountable across distance, and cannot surface the hard choices about whose need gets met. The system stays rooted but remains local.
The real break comes when practitioners must choose: Do I use this methodology or honor this relationship? Do I fund based on evidence or trust? The system fragments because the question itself assumes they’re incompatible. Funders become defensive. Communities withdraw. Resources pool in spaces comfortable with quantification, starving those that aren’t.
Section 3: Solution
Therefore, design the giving commons to hold both instrumental rigor and relational depth as generative constraints, not competing goods.
This pattern reframes effectiveness and value as reciprocal requirements, not tradeoffs. The mechanism works like this: In a healthy commons, rigor serves relationship, and relationship sharpens rigor.
Start from a living systems insight: organisms that survive and adapt hold multiple feedback loops simultaneously. A tree measures water uptake (instrumental) while sensing the soil ecology it depends on (relational). Neither dominates; both inform. Similarly, a giving commons needs both the cooling clarity of evidence and the warming presence of trust.
When a funder asks “What is the cost per outcome?”, that’s legitimate rigor. But the next question must be relational: “Who defined that outcome? Did they participate in measuring it? What are we not seeing?” This isn’t additional work—it’s how you prevent misdirected impact. When a community says “We know what we need,” that’s legitimate authority. But the next move is rigorous: “How do we test whether this approach scales? What would tell us if it’s not working?” This isn’t external imposition—it’s shared accountability.
The pattern activates when practitioners commit to triangulation: using evidence to sharpen relational intuition, and using relational feedback to question what gets measured. This creates what moral philosophy has long called practical wisdom—the capacity to apply principle to particular circumstance with both clarity and care. It requires a cognitive shift: from “effectiveness or value” to “effectiveness of relationships, relationships grounded in evidence.”
The commons becomes more vital because it generates new adaptive capacity. Practitioners learn what rigor looks like when it’s accountable to those affected. Evidence becomes sharper because it’s checked against lived experience. Giving flows toward genuine flourishing rather than proxied good.
Section 4: Implementation
In corporate contexts: Design impact investment committees with equal seats for financial analysts and community representatives. Before approving a fund, run a relational audit: Have those affected by the investment approach been asked what success means to them? Require that cost-per-outcome calculations include columns for “community-defined value” alongside funder-defined metrics. When a company’s impact team proposes scaling a program to ten markets, demand they first slow down and map the relationships already present—existing trust networks, local knowledge holders, governance forms. This prevents the common failure of metric-optimized solutions that disrupt working commons. Use quarterly board reviews to track both: How accurate was our forecast? How much deeper did relationships grow? Neither answer alone is sufficient.
In government contexts: Establish Evidence and Relation Teams working in tandem in policy shops. When designing a public health intervention, the evidence team produces the randomized trial data; the relation team simultaneously conducts deep listening in the affected communities, surfaces tacit knowledge, identifies whose voice is missing from the problem statement. Create a decision gate: No policy scales without both the RCT result and a signed statement from community representatives saying the evidence aligns with what they know and want. This prevents the silencing that happens when government imposes evidence-based solutions communities never consented to. In budget cycles, allocate funding in tranches tied to these paired reviews.
In activist contexts: Build measurement into campaigns from day one, but measure what movements actually care about: not just “signatures collected” but “relationships formed that sustain organizing across campaigns.” Create structured reflection cycles where organizers explicitly ask: Did this tactic build power with those most affected, or did it extract from them? Track both the instrumental win (policy changed) and the relational cost (Did we burn out local leaders? Did we create dependency on external organizers?). Use data to prove what you know intuitively—that the most effective campaigns are the ones where affected communities lead from the start. Let evidence legitimize what relational wisdom already knows.
In tech contexts: Redesign impact measurement systems to surface relational data, not hide it. When a giving platform shows “500,000 people helped,” add a second metric: “How many of those 500,000 people chose which problems got addressed?” or “What percentage of beneficiaries participated in designing the solution?” Use algorithmic transparency to force visibility: Show users not just the cost-effectiveness calculation, but also the stakeholders it excluded. Create feedback loops where beneficiaries can directly contest or contextualize impact claims. When AI systems predict “high-impact” interventions, require them to also flag “relational risks”—interventions that might score high on metrics but violate community autonomy or existing social ties.
Across all contexts: Establish relational-rigor partnerships as the basic unit. Never let evidence live in one silo and relationships in another. Pair your researcher with community co-investigators. Pair your impact manager with affected community members. Compensate both equally; credit both in findings. Create a shared accountability structure where both can veto a decision if evidence and relationship seriously diverge. This isn’t consensus—it’s the kind of creative friction that produces wisdom. When conflict arises (evidence says X works; community knows Y works), don’t resolve it by choosing sides. Investigate the conflict as data. Often, you’ll find evidence was measuring the wrong outcome or community knowledge wasn’t fully articulated. The pattern works when people treat disagreement as a signal to look deeper, not as a problem to eliminate.
Section 5: Consequences
What flourishes: This pattern generates new adaptive capacity in the giving commons. Funders develop the ability to recognize genuine impact versus metric artifacts—they learn to see when a program succeeds on paper but fails in practice. Communities develop the ability to access resources at scale without losing agency—they learn to speak the language of evidence without abandoning their own ways of knowing. Most importantly, practitioners become capable of holding paradox: the evidence is real, and lived experience is real, and sometimes they contradict, and that contradiction teaches. The pattern also renews stakeholder architecture (which scored 3.0) by creating legitimate cross-stakeholder decision-making structures where before there was only top-down metrics or bottom-up resistance.
What risks emerge: The pattern can hollow into performance ritual. Practitioners create the appearance of relational-rigor partnership while power remains unchanged—communities participate in measurement but not in resource decisions. Evidence and relationship become another box to check rather than a generative tension. Watch for this specifically: If the same voices always drive the final decision, you’re not holding both.
Resilience scores low at 3.0 because this pattern requires high-touch coordination. It doesn’t scale through systems or rules; it scales through practitioner maturity. If practitioners lack the skills to hold both without collapsing into one or the other, the pattern fragments. The commons becomes vulnerable to leadership transitions—the moment a new funder arrives who doesn’t believe in this integration, infrastructure crumbles. Guard against this by making relational-rigor partnership a structural requirement (in contracts, in governance), not just a cultural value. Also watch for metric creep: As pressure mounts to prove impact, organizations slowly re-privilege quantification, and relationships become what they measure, not what they honor.
Section 6: Known Uses
GiveWell’s shift toward community voice (2018–present): The effective altruism flagship initially succeeded through ruthless evidence: cost per malaria net distributed, disability-adjusted life years saved. But it encountered a wall. Communities in malaria-endemic regions asked: Why are we not part of designing what “effective” means? GiveWell began structuring recommendation processes differently—adding local researchers and affected community members to evaluation teams, not as consultants but as co-evaluators. When evaluating a maternal health program, they now ask: How did the program community itself define successful maternal health, and does that align with the metric we’re using? This caught cases where high-coverage programs were driving outcomes communities didn’t actually want (e.g., medical delivery when community preferred midwife-attended birth). The instrumental metric was accurate; the relational understanding had been incomplete. By holding both, GiveWell became more credible to affected communities while maintaining rigor.
Open Society Foundations’ Participatory Grantmaking (2016–ongoing): OSF inverts traditional funder authority by allocating 10% of certain program budgets directly to participatory processes where affected communities choose grantees. This might seem purely relational, but it’s paired with rigorous documentation: What assumptions about “good work” did participatory grantees surface? How did their selections differ from expert recommendations, and what did we learn? In criminal justice, for instance, participatory grantees funded grassroots organizations experts had overlooked. Evidence later showed these groups had longer relationship-based impact—communities trusted them, which meant behavior change lasted beyond the grant. The instrumental metric was relationship depth, measured longitudinally. Neither approach alone would have worked; together, they showed that community selection often predicts durability better than expert evaluation.
Participatory Action Research in health (exemplified by partners like Community Engagement Solutions): In global health, practitioners began pairing randomized controlled trials with participatory research cycles. When testing a tuberculosis treatment adherence program, conventional approaches would measure: Did people take pills? Real-world implementation discovered that adherence failed not because the evidence was wrong, but because program designers hadn’t asked people why pill-taking was hard. Participatory research surfaced that treatment schedules conflicted with agricultural cycles, or that stigma made clinic visits dangerous. Evidence gathered with communities rather than about them produced different solutions. The pattern worked because researchers treated community members as co-investigators who could read the data, not as subjects to be measured. Effectiveness improved because it was grounded in relational reality.
Section 7: Cognitive Era
In an age where AI systems predict and optimize impact at scale, this pattern becomes more critical and more difficult. AI can now process millions of data points and identify cost-effective interventions faster than human deliberation allows. The temptation intensifies: Why include relational slowness when the algorithm is more accurate?
But the era also creates new leverage. AI systems can now map and visualize social relationships—network effects, trust flows, knowledge distribution—at scales previously invisible. A tech platform can show, in real time, which interventions are strengthening or eroding community ties. This makes the relational dimension measurable in new ways, though not reducible to simple metrics. An intervention that costs $5 per person helped but fractures relationships that sustain collective action shows up differently when you track relationship strength alongside outcome metrics.
The specific risk: AI will optimize for what’s measurable and discard what’s hard to quantify. Relational damage—the slow erosion of community autonomy, the normalization of surveillance in aid delivery, the replacement of mutual aid with algorithmic allocation—happens silently if you’re only tracking outcomes. The pattern requires that practitioners refuse to let AI systems simplify effectiveness to measurable dimensions alone. Demand that AI systems be trained on relational data, surface relational risks in their recommendations, and remain subordinate to human-relational judgment on questions of dignity and agency.
New opportunity: Use AI to amplify community voice. Instead of experts deciding what counts as impact, train AI systems on community-defined success (gathered through participatory processes) and let the algorithm find patterns in how communities achieve it. This inverts the usual power flow: technology serves relational authority rather than replacing it. The tech context is medium, which is appropriate—AI is neither the solution nor the problem; it’s a tool that amplifies whatever values you embed in it.
Section 8: Vitality
Signs of life:
The pattern is working when you see cognitive discomfort in decision-making—funders and community partners genuinely struggling with cases where evidence and relationship pull in different directions, and treating that struggle as important rather than annoying. When decisions slow down because someone insists on asking a relational question, that’s health, not bureaucracy.
You’ll see evidence changing questions rather than answers—practitioners using data to surface what wasn’t visible (community members aren’t accessing services because of transportation, not because they don’t value them), not to prove predetermined conclusions. When evidence shifts what gets investigated, the pattern is alive.
Most concretely: Affected communities seeking out these funding structures and staying in them. Not because they’re forced to, but because they have more genuine influence than elsewhere. Long-term retention of community partners and repeated participation by beneficiaries in measurement processes is the strongest signal.
Signs of decay:
The pattern is hollow when community participation becomes ceremonial—you have the meetings, but decisions have already been made. Communities sense this quickly and disengage. When relational roles are unpaid while analytical roles are fully staffed, you’ve already chosen which side matters.
Watch for metric proliferation without question-shifting—when you start tracking more and more indicators without asking what assumptions changed in measurement. This signals that evidence is becoming decorative rather than generative.
Most telling: When people stop describing tension as valuable. If you hear “We finally got effectiveness and values aligned!” with relief rather than ongoing creative friction, the pattern has collapsed into false harmony. Real alignment preserves the productive disagreement—it just channels it productively.
When to replant:
Replant this pattern when communities withdraw from decision-making or when funders begin making strategic choices without community input. This often happens during leadership transitions or rapid scaling. The moment you notice that relational representation in decision structures is being treated as a nice-to-have rather than a requirement, redesign. Also replant if you see metric capture—when the measurement system becomes so elaborate that practitioners spend more time feeding it than doing actual work with communities. That’s a sign the pattern has inverted.