Evaluating the effectiveness of outreach within the EA movement

Introduction

The growth of the Effective Altruism movement has largely relied on conferences, local groups, and educational programs to raise awareness of the movement. However, much of the current research suggests these events to be largely ineffective in encouraging participants to engage further with EA communities. This paper examines this by reviewing existing studies by EA groups, as well as analysing data collected during the Leaf 2025 summer course. Ultimately, this paper argues that attitudinal change, as well as behavioural change, is often limited among individuals who engage in brief effective altruism-related events and programmes.

Literature review

Research on outreach by EA organisations indicates that many participants report shifts in their priorities and commitments after engaging with EA, although this does not frequently correspond with behavioural changes. Rethink Priorities’ analysis of the 2019 EA Survey found that 42% of respondents reported changing their primary cause area after becoming involved with an EA community. However, relatively few respondents had actually made career or behavioural changes to align with EA priorities.^[1] This suggests that while individuals may perceive themselves as having been influenced, the measurable effects are often smaller than self-reports imply, linking to a clear intentions-behaviour gap. Furthermore, a 2023 study of EAGx conferences, which compared the attitudes and behaviours of attendees with non-attendees, found no statistically significant differences between the two groups.^[2] On one hand, the Rethink Priorities study suggests that engagement with EA communities is correlated with changes in priorities and values. However, taken together, the studies hint that the measurable impacts of interventions on behaviour are minimal.

Methodology

For the Leaf dataset ^[3], linear regression was used to explore correlations between changes in participant views, course completion levels, and likelihood to recommend the program.

View changes were measured via responses to the question: “Imagine a situation where you intend to do good (e.g., to improve others’ lives or the world) with a certain limited amount of resources available (e.g., your time or money). You can decide how to allocate your resources by choosing from different options that all do good. The stakes are high. In such a situation, when you can choose between different options of doing good you should follow evidence and reason to do what is most effective, even if you emotionally prefer another option.” Responses ranged from “strongly agree” to “strongly disagree”. Responses were converted to a numerical six-point scale for statistical analysis. Course evaluation was measured by participants rating likelihood to recommend the course on a 1–10 scale. Furthermore, the amount of course completed was tracked through combining meetings attended and worksheets submitted.

Paired t tests were also used to compare participants pre and post cause responses to a question asking them to rate the extent to which they would prioritise each of the following cause areas:

• Animal rights or welfare
• Authoritarianism and corruption
• Biodiversity and conservation
• Climate change
• Education and employment
• Equality (gender, racial, LGBTQ+)
• Human rights
• Poverty and health in the UK
• Poverty and health in low-income and middle-income countries.
• Nuclear security
• Pandemic preparedness or other biorisk reduction
• Refugees and immigration
• Risks posed by artificial intelligence

The data consists of the 119 participants of the program who completed the post-survey, limiting the data's statistical reliability. 

Data Analysis

Participants' responses were tracked before and after the Leaf 2025 course to evaluate belief changes. Overall, the dataset revealed very little net change in views towards the statement ‘we should prioritise what is evidenced as best over what we emotionally prefer’. However, there were some changes in individuals' prioritisation of certain cause areas. 

• There was a strong relationship (r = 0.98) between agreement with the view we should prioritise evidence over emotional preference before and after the course, suggesting views before the course to be the best indicator of views after the course. This would evidence that there were very minimal view changes in participants following the course. 
• Among those participants who agreed with the view that we should prioritise evidence over emotional preference in resource allocation, there was a moderately positive correlation (r = 0.50) between the extent of view change and the likelihood of recommending the course. This suggests that the majority of participants were able to view the course as valuable and worthwhile, regardless of whether they experienced any view changes.
• A high correlation (r = 0.98) was observed between the proportion of the course completed by participants and their likelihood to recommend it. This suggests that engagement with the course was far more likely than belief change to influence whether participants found the course worthwhile. 
• Positive shifts were found in terms of participants prioritisation of certain cause areas. Prioritisation of the risks posed by AI moved from an average score of 3.29 to 3.61. Prioritisation of nuclear preparedness and biorisk reduction showed similar trends, with average prioritisation rating increasing from 3.03 to 3.20 and 2.97 to 3.53, respectively. In contrast, prioritisation of human rights dropped from an average score of 4.14 to 3.92. Additionally, average prioritisation of poverty and health in low and middle income countries dropped from 3.95 to 3.77. Thus, cause prioritisation somewhat shifted after the program to align greater with the 8000 most pressing cause areas. However, standard deviations were large for all of these values, making the results less statistically significant. 

Conclusion

Ultimately, while outreach is often perceived as impactful, the data reviewed in this essay suggests it to be largely ineffective in changing individual’s behaviour. However, this does not diminish the value of outreach by EA communities in providing support and guidance for members. The value of outreach in community support can also be observed more broadly. For instance, campaigns to reduce smoking work best not through their ability to persuade individuals, but rather through creating supportive social environments where the desired choices become easier and more visible.^[4]

Similarly, within the realm of EA, research suggests that community initiatives create social connections between members of EA organisations, which have been paramount in reducing value drift amongst members.^[5] Although, further research may be needed to better establish the value of strengthening community within EA groups as a means by which to ensure longer term support from members.

1. ^{^}

   Rethink Priorities. (2019). EA Survey 2019 series: Cause prioritization. Effective Altruism Forum. https://rethinkpriorities.org

2. ^{^}

   Centre for Effective Altruism. (2023). EAGx impact evaluation report. Effective Altruism Forum. https://forum.effectivealtruism.org

3. ^{^}

   Leaf. (2025). Leaf beliefs data analysis [Google Sheets]. https://docs.google.com/spreadsheets/d/1ZozDTOpjUKfQSJ9tz_LlM7KkBaxJjBmGkm3ZghxxAmo

^{^}

Stead, L. F., Carroll, A. J., & Lancaster, T. (2017). Group behaviour therapy programmes for smoking cessation (Cochrane Review, Issue 3). Cochrane Database of Systematic Reviews. https://doi.org/10.1002/14651858.CD001007.pub3

^{^}

Jurczyk, M. (2024). Value drift in the Effective Altruism movement: A qualitative analysis. Effective Thesis. https://www.effectivethesis.org/finished-theses/value-drift

Show all footnotes