In this year’s SBP-BRiMS data challenge, we ask participants to apply state-of-the-art theoretical knowledge and machine learning techniques to predict technology adoption on a small-scale social network dataset that was collected 70 years ago. This dataset, known as the Medical Innovation Network Dataset, was constructed by Ron Burt (1987), based on a classic sociological study of technological innovation, conducted by Coleman, Katz, and Menzel (1957, 1966) in the mid-1950s.
Burt used this dataset, consisting of over 200 practicing physicians across four cities in Illinois, to test two key mechanisms that explain the physicians’ adoption of the newly invented antibiotic, tetracycline. Coleman, Katz, and Menzel proposed in the original study that social cohesion (i.e., direct social connections that physicians had with other physicians who already adopted) was the driver of adoption. In contrast, Burt proposed and demonstrated that, rather than social cohesion, tetracycline adoption was primarily driven by structural equivalence (i.e., social comparison and competition among physicians who were socially similar). Burt’s classic study is a social scientific exemplar of theoretical and methodological advancement in its own right, masterfully demonstrated through making superior behavioral predictions on Coleman, Katz, and Menzel’s classic data. In the same spirit, the current prediction challenge aims to gauge the theoretical, methodological, and computational advances over the past four decades since Burt’s study, using the same Medical Innovation Network Dataset that he used to challenge Coleman, Katz, and Menzel.
The dataset contains information on the tetracycline adoption times of over 100 physicians in Four Illinois cities (Peoria, Bloomington, Quincy, Galesburg), along with their social network ties and additional individual-level attributes.
The goal of this challenge is to make the most accurate prediction on the tetracycline adoption year/month for 24 physicians in the Medical Innovation Network dataset. Challengers are free to use ANY method/model. See detailed description here.
You need to submit 4 things: a 2-page extended abstract in conference format, a poster to be exhibited at the conference, a single PowerPoint teaser slide - that we can use to promote your entry, and the code/prediction output.
Challenge Paper: A 2-page extended abstract describing the procedure, theories and models/ methods used for prediction, performance metrics, and references. Specifically:
Poster: A poster that summarizes the Challenge Paper (above) for presentation at the conference. Posters are not required before the conference. Posters must be 3’ x 4’ brought to the conference, the conference is unable to print posters.
Promotion/teaser Slide: This is a single PowerPoint slide. The purpose of this slide is to excite people to come to your poster. This slide will also be put online. We will use this slide to promote your entry. This slide should contain:
Prediction Output and Code: In addition to the challenge paper, poster, and teaser slide, submit a zip file that contains your prediction code and three output data files. See detailed instructions.
Each entry will be evaluated on the accuracy of the three prediction outcomes (percent correct) separately. (The most accurate on the month/year prediction may not necessarily be the most accurate on the dyadic adoption order.)
Challenge Response Submission: 21-Aug-2025
Author Notification: 4-Sep-2025
Final Files Due: 1-Oct-2025
Submissions will be accepted at the conference installation on Easychair, https://easychair.org/conferences?conf=2025sbpbrims. Please email sbp-brims@andrew.cmu.edu with any questions. Authors must register prior to uploading.
All entries will send at least one team member to SBP-BRiMS2025 who will be registered for the conference by the early registration deadline to represent their entry.