匿名資助提案以減少偏見

2021 年，美國祇有 11 所大學獲得了所有授予大學和學院的聯邦研究經費的 25%。 。

Despite receiving 25% of all federal research funding awarded to universities and colleges in 2021, just 11 U.S. universities are responsible for distributing the funds. Of these top research-funded universities, nine of them are located on either the East or West Coast. Only University of Michigan and University of Pittsburgh are representing the middle of the country. This trend has been consistent since 2010.

儘管 2021 年授予大學和學院的聯邦研究經費有 25%，但只有 11 所美國大學負責分配這些資金。在這些頂尖的研究資助大學中，有九所位於東岸或西岸。只有密西根大學和匹茲堡大學代表了中部地區。這一趨勢自2010年以來一直保持一致。

However, there are strong research programs at universities throughout the U.S. So why does such a small set of universities receive such a large portion of federal research funding? Is there bias toward these institutions or a hesitancy to criticize proposals from elite universities? Whatever the cause may be, the problem extends beyond federal research funding and also affects foundations that support research.

然而，美國各地的大學都有強大的研究項目，那麼為什麼這麼一小部分大學卻獲得如此大一部分的聯邦研究經費呢？是否對這些機構有偏見，或對批評精英大學的提案猶豫不決？無論原因是什麼，這個問題超出了聯邦研究經費的範圍，也影響到支持研究的基金會。

I’m the executive director of the Arnold and Mabel Beckman Foundation, where we provide basic science research grants with the vision of supporting young scientists today for tomorrow’s breakthrough discoveries. Our mission is to support “young scientists, and especially those that don’t yet have the clout to win major research grants.” One of my main objectives is to ensure that we solicit and evaluate all proposals using an inclusive, fair, and unbiased process. The Beckman Foundation’s success depends on our ability to identify and fund new exciting and innovative research ideas, and we believe that biases in our processes can impede us in accomplishing our mission.

我是阿諾德和梅布爾·貝克曼基金會的執行董事，該基金會提供基礎科學研究資助，其願景是支持今天的年輕科學家實現明天的突破性發現。我們的使命是支持「年輕科學家，特別是那些尚未具備贏得重大研究資助的影響力的科學家」。我的主要目標之一是確保我們採用包容、公平和公正的流程來徵求和評估所有提案。貝克曼基金會的成功取決於我們識別和資助新的令人興奮和創新的研究想法的能力，我們相信我們流程中的偏見可能會阻礙我們完成使命。

Like many basic science funding organizations, we rely on peer review to evaluate and recommend the research projects that we consider for funding. One challenge in the basic sciences is the concentration of grant funding at relatively few U.S. institutions, and my colleagues and I have been concerned that our traditional peer review processes may be perpetuating this concentration.

與許多基礎科學資助組織一樣，我們依靠同儕審查來評估和推薦我們考慮資助的研究計畫。基礎科學面臨的一個挑戰是撥款集中在相對較少的美國機構，我和我的同事一直擔心我們傳統的同儕審查程序可能會延續這種集中。

Admittedly there are many benefits to continue funding established institutions that have made the investments in the infrastructure, research teams, and specialized equipment to conduct advanced research. But this also raises some concerns: Are we stifling research progress when only those with the access and connections to these elite institutions are able to participate? How many great ideas and breakthroughs do we ignore because they aren’t from the “traditional” places? Do the well-resourced institutions become less innovative if they get an advantage through our peer-review system for continued funding?

誠然，繼續資助那些在基礎設施、研究團隊和專業設備上進行投資以進行高級研究的知名機構有很多好處。但這也引起了一些擔憂：當只有那些有權訪問這些精英機構並與其有聯繫的人才能參與時，我們是否會扼殺研究進展？有多少偉大的想法和突破被我們忽視，因為它們不是來自「傳統」地方？如果資源充足的機構透過我們的同儕審查系統獲得持續資助的優勢，它們的創新能力是否會降低？

Of course, as a foundation focused on basic science, we turned to the data to examine the evidence and test our hypothesis. Starting in 2019, we did a detailed analysis of our own grant awards at the Beckman Foundation, and we saw a similar concentration of funding in our awards to the same institutions year after year. We asked ourselves: Are we identifying and funding the best scientific ideas, or does the institutional affiliation of an applicant unduly factor into our review processes?

當然，作為一個專注於基礎科學的基金會，我們轉向數據來檢驗證據並檢驗我們的假設。從 2019 年開始，我們對貝克曼基金會的資助金額進行了詳細分析，發現我們的資助金額年復一年地集中在同一機構。我們問自己：我們是否正在識別和資助最好的科學想法，或者申請人的機構隸屬關係是否不適當地影響到我們的審查過程？

This question led us change the proposal review process we used for our 2020 applications to a dual-anonymized structure to investigate if we saw any implicit bias in our reviews towards prestigious institutions. We then tracked our application statistics for the next four years of applications and saw a shift away from the prestigious institutions that enabled more applications from diverse universities to advance in our selection process. Our complete study methodology and findings are published in a 2024 eLife article.

這個問題促使我們將 2020 年申請所使用的提案審查流程改為雙匿名結構，以調查我們的審查中是否發現對知名機構有任何隱性偏見。然後，我們追蹤了未來四年的申請統計數據，發現著名機構的轉變使得更多來自不同大學的申請能夠在我們的選拔過程中取得進展。我們完整的研究方法和結果發表在 2024 年 eLife 文章中。

In implementing this change, we required our applicants themselves to remove any gender, race, or institutional affiliation identifiers from their proposals. The reviewers were only provided with the anonymized research proposal without any supporting information about the applicant themselves.

在實施這項變更時，我們要求申請人本人從其提案中刪除任何性別、種族或機構隸屬關係識別碼。審查者只收到匿名的研究計劃，沒有任何關於申請人本身的支持資訊。

Interestingly, we found a reduction in the relative advantage of applicants from the more prestigious institutions to advance in our review process after the anonymization step. Before anonymization, the proposals from elite institutions had 1.5 times greater-than-average chance of being selected to advance for further consideration, and after anonymization this advantage to advance dropped to only 1.2 times the average.

有趣的是，我們發現，在匿名步驟之後，來自較有聲望的機構的申請人在進入我們的審核過程中的相對優勢有所下降。在匿名化之前，來自精英機構的提案被選中進入進一步考慮的機會是平均水平的1.5倍，而在匿名化之後，這一優勢下降到只有平均水平的1.2倍。

This shift in which proposals advance in our reviews brings forward more excellent scientific ideas from those not at these elite institutions to be included in the final stages of our review process to be considered for funding. The anonymization resulted in a reduction, but not elimination, of the advantage that the prestigious institutions had in our review process, likely indicating that the prestigious institutions do have an intrinsic advantage: Their researchers and resources put forward above-average and compelling research proposals. However, our finding that the likelihood for these proposals to advance was reduced showed us that these advantages of being associated with certain prestigious institutions were given undue influence in our review processes.

這種在我們的審查中提出提案的轉變帶來了來自這些精英機構以外的人的更多優秀的科學想法，這些想法將被納入我們審查過程的最後階段，以考慮資助。匿名化導致知名機構在我們的審查過程中所具有的優勢減少，但並沒有消除，這可能表明知名機構確實具有內在優勢：他們的研究人員和資源提出了高於平均水平且令人信服的研究建議。然而，我們發現這些提案獲得推進的可能性降低，這表明與某些著名機構合作的這些優勢在我們的審查過程中受到了不當影響。

Ensuring that our review processes are fair and inclusive to all of our applicants regardless of their institution was the goal of our process change, but we also realized the additional benefits that the anonymized proposals were easier to read and evaluate, which reduces reviewer fatigue. During the review discussions we stayed focused on the innovation and the science in the proposal without additional discussions of perceived mentor status, publication rates, journal impact factors, and other ancillary topics.

確保我們的審核流程對所有申請人（無論其機構如何）都是公平和包容的，這是我們流程變更的目標，但我們也意​​識到匿名提案更容易閱讀和評估的額外好處，從而減少了審核者的疲勞。在評審討論期間，我們始終專注於提案中的創新和科學，沒有額外討論感知的導師地位、發表率、期刊影響因素和其他輔助主題。

Like all big systemic challenges, there are many possible solutions being explored. While we have focused on the dual-anonymized review approach, other

與所有重大系統性挑戰一樣，我們正在探索許多可能的解決方案。雖然我們專注於雙重匿名審查方法，但其他