How to Ask for a Favor: A Case Study on the Success of Altruistic Requests

Tim Althoff, Cristian Danescu-Niculescu-Mizil, Dan Jurafsky

Introduction

What motivates a stranger on the internet to buy you a pizza? In most online transactions, value is exchanged for a product or reward. However, in "altruistic" communities, individuals give with no expectation of return. This paper investigates the dynamics of such requests by analyzing Reddit’s "Random Acts of Pizza" (RAOP) community. By studying a platform where every request is for the same item—a pizza—researchers were able to isolate the linguistic and social factors that drive successful persuasion.

Methodology

The researchers analyzed 21,577 posts from RAOP between 2010 and 2013, alongside 1.87 million user submissions across Reddit to determine social status. They employed a multi-dimensional framework:

  1. Narrative Extraction: Using Non-negative Matrix Factorization (NMF), they identified recurring themes in requests. These were refined into concise lexicons representing different narratives such as Money, Job, Student, Family, and Craving.
  2. Linguistic Analysis: The team measured politeness, evidentiality (use of images), reciprocity (promises to "pay it forward"), and sentiment.
  3. Social and Temporal Factors: User "karma" (status), account age, and posting history were tracked, alongside time-of-day and seasonal variables.
Extracted narratives with their associated terms and success rates (SR).

Key Insights

Need vs. Desire

The data reveals that narrative choice is a primary driver of success. Requests centered on "Money" or "Job" issues saw success rates of approximately 32%, whereas those based on "Craving" or "Friends" hovered between 17% and 20%. Donors are significantly more inclined to help those in genuine need rather than those simply seeking a free treat.

The Role of Social Status

Social capital, measured through Reddit's "karma" system, plays a crucial role. Users with higher status within the broader Reddit community are much more likely to have their requests fulfilled. This relationship holds true across all narrative types, with higher status consistently correlating with a higher estimated probability of success.

Estimated probability of success by status decile and narrative type.

Evidence and Reciprocity

"Evidentiality"—providing proof such as a photo of an empty fridge—was the second strongest predictor of success. Additionally, promising "generalized reciprocity" (paying it forward to the community rather than the specific donor) significantly boosted success rates. Interestingly, while gratitude ("thanks in advance") helped, general linguistic politeness had a negligible effect.

Results and Conclusion

The authors' integrated model achieved an AUC of 0.672, significantly outperforming text-only baselines. To illustrate the impact of these factors: a short, 50-word "craving" request from a median-status user has a 9.8% success probability. By contrast, a 150-word "need" request that includes an image, expresses gratitude, and promises reciprocity from a high-status user jumps to a 56.8% success rate.

While these findings are specific to the RAOP community, they offer valuable insights into the mechanics of online altruism and the linguistic strategies that foster prosocial behavior in digital spaces.

Summary updated

BibTeX
@inproceedings{althoff2014how,
  author = {Tim Althoff and Cristian Danescu-Niculescu-Mizil and Dan Jurafsky},
  title = {{How to Ask for a Favor: A Case Study on the Success of Altruistic Requests}},
  booktitle = {ICWSM},
  year = {2014},
}
Figure