Mitra, Tanushree, and Eric Gilbert. “The language that gets people to give: Phrases that predict success on Kickstarter.” Proceedings of the 17th ACM conference on Computer supported cooperative work & social computing. ACM, 2014.
Summary
The language that gets people to give is a fascinating study that attempts to answer two questions – can we predict what crowdfunding campaigns get funded, and what features of the campaign determine its success. Analyzing over 45K Kickstarter campaigns, Mitra et al. build a penalized regression model with 59 control features such as project goal, duration, number of pledge levels etc. Using this as a baseline, they built another model with textual features extracted from the project description. To ensure generalizability, they only chose words and phrases that appear in all 13 categories of Kickstarter campaigns. The control-only model has an error rate of roughly 17%. The use of language features (~20K phrases) reduces the error rate to 2.4%, indicating a non-random increase in accuracy. The paper then relates the top features for both the funded and not-funded cases to social psychology and the theories of persuasion. Of these, campaigns that display reciprocity (the tendency to return a favor), scarcity (limited availability of the product), social proof (others like the product too), authority (an expert designing or praising the product) or sentiment (how positive is the description) tend to be funded more.
Reflection
An exciting aspect of this paper is the marriage of social psychology, statistical modeling, and natural language processing. The authors address a challenging question about what features, language or otherwise, encourage users to invest in a campaign. The paper borrows heavily from theories of persuasion to describe the effects of certain linguistic features. While project features like the number of pledge levels are positively correlated with increased chances of funding, I was surprised to see phrases such as “used in a” or “project will be” influencing successful funding. I am equally interested in how these phrases relate to specific aspects of persuasion – in this case, reciprocity and liking/authority. The same phrases can be used in different contexts to imply different meanings. I am curious to know if the subjectivity index [1] of project descriptions make any contribution to a fund or no-fund decision.
I would expect that another important aspect of successful campaigns would be the usefulness of a product to the average user. While this is hard to measure objectively, I was surprised to find no reference to this in any of the top predictors. Substantial research in sales and marketing seems to indicate a growing emphasis on product design for successful marketing campaigns [2].
A final aspect that I find intriguing is the deliberate choice of treating all products equally on Kickstarter. How valid is this assumption when one considers funding a documentary vs. earphones? It is likely that one might focus much more on contents and vivid descriptions, while the other might focus more on technical features and benchmarks?
The paper throws open the entire field of social psychology and offers a great starting point for me to read and understand the interplay of psychology and linguistics.
Questions
- Do different categories of campaigns experience different funding patterns?
- Are certain types of projects more likely to be funded as compared to others?
- While social psychology is an important aspect of successful campaigns, perhaps it would make sense only in conjecture with what the product really is?
[1] Theresa Wilson, Janyce Wiebe, and Paul Hoffmann (2005). Recognizing Contextual Polarity in Phrase-Level Sentiment Analysis. Proc. of HLT-EMNLP-2005.
[2] Why the Product is the Most Important Part of the Marketing Mix. http://bxtvisuals.com/product-important-part-marketing-mix/
