2025 Einstein Foundation Award Winners' Colloquium

What makes a discipline scientific? How can a discipline recover from a crisis of credibility, and bolster its trustworthiness? Psychology's replication crisis and credibility revolution provide a valuable case study. I review psychology's response to its crisis, and draw lessons for other disciplines that may face similar crises. Psychology's crisis was triggered by failures to replicate, fraud, and questionable research and publication practices. The most important avenues for improving the integrity and credibility of our field have been: increasing transparency, improving journals and peer review, and supporting post publication critique. Without attention to these fundamental issues, the 'self-correcting mechanisms' of science, including meta-analysis, cannot save a discipline.

Concerns over the reproducibility of published research have grown in many research fields, but empirical data to inform policy in local contexts are still scarce. The Brazilian Reproducibility Initiative was set up to fill this gap by performing a multicenter replication of published experiments from Brazil, using three common experimental methods in lab biology. A total of 56 laboratories performed 143 replications of 56 experiments, with replication rates varying between 20 and 44% according to predefined criteria. Moreover, examining protocol deviations, technical issues, data handling and human resources in participant labs provide a fascinating glimpse of factors that limit reproducibility in Brazilian biomedical science. The project has given rise to the founding of the Brazilian Reproducibility Network, which now works to foster reproducible research in the country through community building, training, meta-research and advocacy at a national level in diverse research fields.

High-throughput sequencing data has become the backbone of modern biomedical research, but its reliability is often challenged by batch effects and technical artifacts. By systematically introducing errors during wet lab sequencing workflows, we aim to measure their impact on downstream analysis using our established machine learning tool for quality prediction. Building on prior work in functional genomics and in the concept of Quality Imbalance, we aim to bridge bioinformatics and experimental design to improve reproducibility and data interpretation in real-world research settings.