This project aims to understand the downstream effects of different types of content posted on online platforms. This involves running a field experiment in which different types of content are posted on an online platform and the effects of posting such content, e.g. user engagement and follow-on content production, are then examined.
Intervention
The intervention consists of posting comments of different categories on an online social media platform.
Experimental design
As users make posts on the online platform, new posts are randomized into different treatment groups, each of which receives a comment responding to their post.
Design details
Randomization method
Randomization is done via an automated script.
Randomization unit
Individual
Clusters
N/A.
Observations
A minimum of 3000 users. If the infrastructure to conduct the experiment remains stable (in terms of how long it remains feasible to keep posting treatment comments at scale), I will expand the sample to include up to approximately 12,000 users. While these are the number of observations planned, in cases where some of the treated users may close their accounts or there are issues with posting comments at scale, the actual minimum number in the sample may be lower.
Arms
Individual users will be randomized into treatment groups with equal probabilities.
Primary outcomes
The primary outcomes are:
1) Short term content production, i.e. replies received on the comments posted as part of the treatment.
2) Longer term content production, i.e. content produced by treated users following the treatment.
Secondary outcomes
1. Engagement received on treatment comments, such as their virality as measured by comment scores.
2. Key differences between treatment comment categories based on analysis of a) the language (e.g. sentiment and other content attributes) of the comments posted and b) the language of the replies they get.
Cited by
No citation data available for this registration yet.