Search

Machine learning (ML) methods are proliferating in scientific research. However, the adoption of these methods has been accompanied by failures of validity, reproducibility, and generalizability. These failures can hinder scientific progress, lead to false consensus around invalid claims, and undermine the credibility of ML-based science. ML methods are often applied and fail in similar ways across disciplines. Motivated by this observation, our goal is to provide clear reporting standards for ML-based science. Drawing from an extensive review of past literature, we present the REFORMS checklist ($\textbf{Re}$porting Standards $\textbf{For}$ $\textbf{M}$achine Learning Based $\textbf{S}$cience). It consists of 32 questions and a paired set of guidelines. REFORMS was developed based on a consensus of 19 researchers across computer science, data science, mathematics, social sciences, and biomedical sciences. REFORMS can serve as a resource for researchers when designing and implementing a study, for referees when reviewing papers, and for journals when enforcing standards for transparency and reproducibility.

Erowid.org is a website dedicated to documenting information about psychoactive substances, with over 36,000 user-submitted drug Experience Reports. We study the potential of these reports to provide information about characteristic experiences with drugs. First, we assess different kinds of drug experiences, such as 'addiction' or 'bad trips'. We quantitatively analyze how such experiences are related to substances and user variables. Furthermore, we classify positive and negative experiences as well as reported addiction using information about the consumer, substance, context and location of the drug experience. While variables based only on objective characteristics yield poor predictive performance for subjective experiences, we find subjective user reports can help to identify new patterns and impact factors on drug experiences. In particular, we found a positive association between addiction experiences and dextromethorphan, a substance with largely unknown withdrawal effects. Our research can help to gain a deeper sociological understanding of drug consumption and to identify relationships which may have clinical relevance. Moreover, it can show how non-mainstream social media platforms can be utilized to study characteristics of human behavior and how this can be done in an ethical way in collaboration with the platform providers.

We present the first full description of Media Cloud, an open source platform based on crawling hyperlink structure in operation for over 10 years, that for many uses will be the best way to collect data for studying the media ecosystem on the open web. We document the key choices behind what data Media Cloud collects and stores, how it processes and organizes these data, and its open API access as well as user-facing tools. We also highlight the strengths and limitations of the Media Cloud collection strategy compared to relevant alternatives. We give an overview two sample datasets generated using Media Cloud and discuss how researchers can use the platform to create their own datasets., Comment: 15 pages, 9 figures, accepted (minus the 3-page, 3-image appendix given here) for publication and forthcoming in Proceedings of the Fifteenth International AAAI Conference on Web and Social Media (ICWSM-2021)

We present a study to detect friendship, its strength, and its change from smartphone location data collectedamong members of a fraternity. We extract a rich set of co-location features and build classifiers that detectfriendships and close friendship at 30% above a random baseline. We design cross-validation schema to testour model performance in specific application settings, finding it robust to seeing new dyads and to temporalvariance.

"All models are wrong, but some are useful", wrote George E. P. Box (1979). Machine learning has focused on the usefulness of probability models for prediction in social systems, but is only now coming to grips with the ways in which these models are wrong---and the consequences of those shortcomings. This paper attempts a comprehensive, structured overview of the specific conceptual, procedural, and statistical limitations of models in machine learning when applied to society. Machine learning modelers themselves can use the described hierarchy to identify possible failure points and think through how to address them, and consumers of machine learning models can know what to question when confronted with the decision about if, where, and how to apply machine learning. The limitations go from commitments inherent in quantification itself, through to showing how unmodeled dependencies can lead to cross-validation being overly optimistic as a way of assessing model performance., Comment: 68 pages, 7 figures

Preprocessors support the diversification of software products with #ifdefs, but also require additional effort from developers to maintain and understand variable code. We conjecture that #ifdefs cause developers to produce more vulnerable code because they are required to reason about multiple features simultaneously and maintain complex mental models of dependencies of configurable code. We extracted a variational call graph across all configurations of the Linux kernel, and used configuration complexity metrics to compare vulnerable and non-vulnerable functions considering their vulnerability history. Our goal was to learn about whether we can observe a measurable influence of configuration complexity on the occurrence of vulnerabilities. Our results suggest, among others, that vulnerable functions have higher variability than non-vulnerable ones and are also constrained by fewer configuration options. This suggests that developers are inclined to notice functions appear in frequently-compiled product variants. We aim to raise developers' awareness to address variability more systematically, since configuration complexity is an important, but often ignored aspect of software product lines.