A family foundation’s uncertainty charter: evaluating ambitious grants without profiling donors

The finding that should change the committee’s question

A family office may be tempted to infer philanthropic preferences from its investment questionnaire: an entrepreneur comfortable with venture capital must prefer moonshot research, while a conservative investor must favor established service delivery. Costa-Gomes and Schoenegger’s study does not support treating that inference as established. It recruited 1,506 UK participants through Prolific, measured several individual differences, and observed incentivized choices involving real charities. Across its main comparisons, measured risk and ambiguity attitudes generally did not explain choices between the authors’ more reliable and more uncertain charity categories. Exploratory equivalence tests and Bayesian analyses supplied additional support for small or absent effects for several measures within this design.

That result is a reason to ask donors directly about philanthropic purposes, not a reason to disregard uncertainty. The study concerns predictors of donor behavior; it does not estimate which charities produce the greatest social benefit or prescribe a grant portfolio. A risky charitable intervention can be attractive on mission grounds even when its probability of success cannot be credibly estimated. Conversely, a persuasive description of very large potential benefits does not establish that the organization can deliver them. The committee needs two separate records: what the family hopes to achieve, and what the available evidence supports about the proposed intervention.

Read the design before using the result

The experiment used three charities in each category. The relatively reliable group included deworming, direct cash transfers and insecticide-treated bed nets; the probabilistic group concerned epidemic prevention, nuclear-war prevention and artificial-intelligence risk. These are the study’s categories and historical selections, not current recommendations from this journal. Even the authors’ “sure-thing” category retained risk. The labels should therefore never be copied onto a live grant memorandum as promises of guaranteed impact.

In Main Choice, participants could give to one of two charities or keep their endowment. In Final Choice, they could give to a single randomly selected charity. The authors explicitly favor interpretation of Final Choice because paired choices can obscure indifference between different combinations of recipient and amount, and excluding non-donors can affect inference. Final Choice occurred later in the experiment, so order effects cannot be ruled out. Some measures of altruistic donor type predicted giving to probabilistic charities in that setting; this does not justify assigning family members fixed psychological labels or screening them for access to ambitious grants.

The study’s recruitment frame was a UK online sample, with representativeness sought on age, sex and ethnicity, not a sample of family offices or major philanthropists. Stakes were small and the donation task differed from multiyear, advised grant decisions. More participants declined to donate than expected, leaving the expected-value information and context-free treatment analyses inconclusive. It would be incorrect to claim that teaching expected-value reasoning was proven ineffective. The practical protocol below is an editorial proposal motivated by these boundaries, not an empirically validated family-office intervention.

The paper also discloses that Philipp Schoenegger received funding from the Forethought Foundation and the Centre for Effective Altruism and held a Global Priorities Fellowship; the selected charities came from Giving What We Can recommendations, which the disclosure identifies with the Centre for Effective Altruism. The authors report that funders had no role in the research or publication decisions. This disclosed relationship should remain visible, and the six selected charities should not be treated as a representative sample of all philanthropy or as independently validated investments.

Replace personality shortcuts with an uncertainty register

Before comparing proposals, the philanthropy director should convene the authorized family decision-makers to record mission priorities, time horizon, acceptable failure, and the distinction between support for services and support for learning. Ask each person what evidence would make a grant worthwhile even if its ultimate goal were not achieved. Record disagreement instead of compressing it into a single risk score. Do not reuse investment suitability or personal psychological data as a substitute for this conversation.

For each proposal, separate three uncertainties. Event uncertainty concerns whether the relevant need or adverse event occurs. Intervention uncertainty concerns whether the proposed activity changes that event or its consequences. Execution uncertainty concerns whether this organization can conduct the activity competently and ethically. These categories are the companion’s operational extension of the paper’s distinction between quantifiable risk and less quantifiable ambiguity. Each entry should name an evidence source, date, assumption owner and plausible disconfirming observation. Where probabilities cannot be defended, record ranges or qualitative uncertainty and explain why; do not fabricate precise numbers to populate a spreadsheet.

A hypothetical epidemic-prevention grant might have uncertain ultimate population outcomes yet a verifiable near-term objective such as completing a specified surveillance-validation study. The committee should keep those different claims separate. Completion of the study is evidence of an activity and perhaps learning; it is not proof that a future epidemic was prevented. A familiar service-delivery grant likewise needs records of reach, quality and additionality. Familiarity must not excuse weak evidence.

Assess proposals alone before deliberating about a portfolio

Because the paper identifies interpretive difficulties in paired choice, a useful committee discipline is to review each grant against a common mission and evidence template before placing proposals side by side. This is a suggested safeguard, not a claim that the experiment proved this meeting procedure superior. Reviewers should independently document the intervention’s theory of change, credible alternatives, execution capacity, uncertainties and reasons to decline. Only after these records exist should the committee compare opportunity costs and dependencies across proposals.

The grants operations lead should preserve the standalone assessments, the subsequent comparative discussion, and any changed judgments. A revision is not automatically a bias: comparison may reveal a genuine tradeoff. The record should distinguish new evidence from rhetorical contrast, enthusiasm for a charismatic founder, or a desire to avoid disagreeing with a senior family member. Measure missing evidence fields and unexplained changes rather than rewarding the volume of grants approved. The committee chair should pause a vote when a material objection lacks an answer or an accountable owner.

Stage learning and protect the ability to stop

For an uncertain intervention, the finance lead and program director can propose a bounded initial commitment linked to an identifiable learning milestone. State the maximum exposure the foundation is prepared to lose, the conditions under which a further tranche may be considered, and the person authorized to release it. This article supplies no allocation percentage: capacity, existing commitments, mission and legal duties differ across families. A staged structure is also not always suitable; it can impose avoidable uncertainty on a grantee or underfund work that needs a continuous runway.

Milestones should test the project’s actual causal argument. A research grant might require a documented feasibility result or an independently reviewed dataset. A delivery grant might require evidence that the intended population receives the service at an acceptable quality. Neither publicity, fundraising momentum nor donor satisfaction alone is a substitute for those project-specific tests. Preserve negative findings and unsuccessful pilots so that the foundation can stop repeating a failed approach.

Precommit to stopping or redesigning when the intervention’s central assumption is contradicted, required data remain inaccessible, safety or integrity concerns arise, or the organization cannot meet a material execution obligation. A missed milestone should trigger a documented review rather than an automatic punitive cancellation: the committee must determine whether the failure reflects learning, a recoverable implementation issue or a broken theory of change. New money requires an explicit decision; enthusiasm should not silently renew the commitment.

Pilot the governance process without claiming an impact result

Apply the template to a small upcoming set of grants and review the process after one decision cycle. The evaluation owner should count completed independent assessments, documented uncertainty statements, resolved evidence requests, reasons for continuation and reasons for stopping. Track staff and grantee burden as well. These are process measures, not evidence of increased charitable impact. A before-and-after change in approval rates would be confounded by the proposals and decision-makers involved and should not be presented as a causal validation.

The original article’s author order was corrected by PLOS on July 22, 2024: Miguel Costa-Gomes precedes Philipp Schoenegger. The source edition and this companion follow that notice while retaining the original 2022 DOI and publication date. The correction concerns attribution and is not an additional research study. This publication is AI-assisted editorial analysis, not individualized investment advice, a new experiment, source-author endorsement or documented independent human peer review. Its practical ambition is narrower and testable: make the reasons for philanthropic commitments sufficiently explicit that the family can learn, disagree responsibly and stop when the evidence warrants it.

Source, attribution and limits

[1] Miguel Costa-Gomes; Philipp Schoenegger. Sure-thing vs. probabilistic charitable giving: Experimental evidence on the role of individual differences in risky and ambiguous charitable decision-making. PLOS ONE (2022-09-22). https://doi.org/10.1371/journal.pone.0273971. CC BY 4.0.

UK Prolific online sample with small monetary stakes; not family-office evidence. Recruitment n=1,506, with exclusions and analysis-specific samples. The authors identify paired-choice indifference/selection problems, possible Final Choice order effects, inconclusive underpowered secondary treatments, exploratory equivalence/Bayesian analyses and limited generalizability. No causal conclusion that financial-risk profiling determines philanthropic preferences, no optimal grant allocation, and no measured advantage of the companion protocol. Corrected author order follows the 2024 publisher notice despite stale original metadata. The disclosed Forethought Foundation/Centre for Effective Altruism funding and fellowship relationship overlaps with the provenance of the charity recommendation set; funders are reported to have had no role in research or publication decisions.

Prepared by the Family Office Private Wealth Journal editorial research desk with AI assistance at the direction of Dr. Freedom Cheteni. Original research authors retain full credit and are not authors or endorsers of this companion. This proposed process has not been evaluated in a new empirical study; no independent peer-review acceptance is claimed. This is educational analysis, not individualized investment, legal or tax advice.

Implementation record

Open the cited sources

1. [1] Miguel Costa-Gomes; Philipp Schoenegger. Sure-thing vs. probabilistic charitable giving: Experimental evidence on the role of individual differences in risky and ambiguous charitable decision-making. PLOS ONE (2022-09-22). ↗