Reproducibility in an AI world

Abstract

Artificial intelligence usage as a scientific tool is increasing in the social sciences. With the earlier “replication crisis” still in mind, are we heading towards more irreproducible research? I will discuss approaches and solutions on how social scientists can structure and document research with AI tools. 

Aims & Scope of META-REP

The behavioral, social, and cognitive sciences are in the midst of an intensive debate about the replicability of their empirical findings. Shaken by the results of many replication projects that have been conducted in recent years, scholars have begun discussing what replicability actually means and when a replication can be regarded as successful vs. failed, whether low replication rates are caused by too many false positive findings in the literature, methodological weaknesses in the replication studies, measurement issues, and/or the underappreciated influence of contextual effects, and what can be done to effectively and sustainably secure a high level of replicability in the behavioral, social, and cognitive sciences.