Despite peer-review and editorial screening, science papers can still contain images or other data of concern. A visual scan of 20,000 papers published in 40 biomedical journals showed that 4% contained inappropriately duplicated images. Papers containing incorrect or even falsified data could lead to wasted time and money spent by other researchers trying to reproduce those results. Thorough image screening before publication would be beneficial for editors, publishers, and readers, and act as a deterrent for fraudulent submissions. There is a great need for high-throughput computational tools to find image duplications and manipulations in scientific manuscripts, and to help detect the growing number of fabricated manuscripts produced by paper mills. Elisabeth Bik will present some case examples, insights, and challenges that she has encountered as a human visual duplication detector.

Dr. Lauer, the NIH Deputy Director for Extramural Research, will discuss one funding agency’s perspectives on the roles and responsibilities of different stakeholders in addressing varying types of research and professional misconduct.

Advances in machine learning technologies have led to the rapid proliferation of automated and semi-automated media (image, video, text, audio) manipulation and synthesis technologies. These technologies make it easier for unskilled individuals to create compelling media manipulations and may reduce our trust in many forms of information. DARPA has made significant investments in the development of media forensics tools that can help defend against such falsified media. This talk will discuss trends in media falsification technologies and the work that has been developed by DARPA programs to defend against falsified media.

The PubPeer website, dedicated to facilitating public discussion of the scientific literature, launched in late 2012. Using open metadata, it creates a dynamic page for every publication with a Digital Object Identifier (DOI) or arXiv ID. Users can post and read comments about each publication. This short-circuits the many potentially conflicted actors stifling the correction of science: authors, journals and institutions. By offering strong anonymity, regulated by strict content guidelines and moderation, PubPeer has created a protected space where even serious criticism could be aired without risk of professional or legal reprisals. The site has helped reveal an unsuspected volume of research misconduct. Users can be alerted to discussion about articles as they work by installing the PubPeer browser and Zotero plugins (which you should all do). Taking stock after 8 years of operation, we stand by our sometimes-controversial decision to allow regulated anonymous commenting, which we believe has reinforced what we see as the primary function of PubPeer - to serve as an early-warning system for readers and users of articles. Official procedures for correcting science remain too slow, unreliable and inefficient; this was especially apparent during the pandemic. We aim to continue integrating sources of high-quality discussion, with recent innovations including comments from preprint servers, scientific Twitter and a sister site for overlay journals called Peeriodicals. Journals and institutions can receive tailored alerts via "PubPeer Dashboards". Our guiding philosophy for the future is to accelerate scientific progress by facilitating rapid, effective and public exchange of discussion about publications, while eschewing all metrics and focusing on the substance of scientific publications.

The COVID-19 pandemic has attracted significant attention to scientific matters related to the cause, treatment, and prevention of the disease that has upended our lives. Alarmingly, not all of the information available on the Internet is what it appears to be. Deceptive memes, bogus ads, and fabricated infographics are proliferating, with all threatening to undermine the public's trust in science. Given the vast scale of the problem, an automated capability that can identify new instances of visual disinformation, trace its origin, and ultimately flag it as being problematic is needed. But compared to text, visual content presents unique challenges for media forensics. This talk presents an end-to-end processing pipeline for image provenance analysis, which works at real-world scale. It employs a cutting-edge image filtering solution that is able to find related images, as well as novel techniques for obtaining a provenance graph that expresses how the images, as nodes, are ancestrally connected. Building from provenance analysis, the talk goes on to introduce a scalable automated visual recognition pipeline for discovering meme genres of diverse appearance. This pipeline can ingest meme images from a social network, apply computer vision-based techniques to extract features and index new images into a database, and then organize the memes into related genres. Recent examples of visual disinformation targeting science will be highlighted, including repurposed imagery, parasitic advertising, and pandemic-related memes. Finally, the talk will conclude with thoughts on continued research in this direction.

Agnotology – a new field dedicated to the study of ignorance -- has provided crucial insights into the corporate staging and management of scientific controversies on topics of great public importance ranging from the dangers of tobacco smoking to global warming, and others in between. (The book and eponymous film Merchants of Doubts or the work of Robert Proctor are examples). A shared goal of these corporate interventions, the agnotologists have argued, is the questioning of scientific evidence and claims so as to create the appearance that these issues are not matters of remarkably solid consensus but, to the contrary, the focus of substantial controversies among scientists themselves. The effect of such misinformation campaigns has been to splinter or confuse the public’s response, dangerously delaying urgent regulatory interventions. But while the agnotologists’ goal has been to study the production of non-knowledge or the destruction of knowledge -- from the feeding of false beliefs to the public to actively creating gaps in the availability of knowledge by making it secret -- their most original and interesting contribution has been the thick description of the techniques, strategies, and rhetoric of the production of doubt about science and medicine – something that is exceptionally effective precisely because it is neither truth nor falsehood, neither knowledge nor ignorance. Its power is its elusiveness. Adding new examples to those exposed by the agnotologists, I argue that the production of doubt is not the production of ignorance but an essentially different regime that, while coming to life in the traditional form of scientific controversies, it does not aim at producing or manipulating the content of knowledge but at strategically subverting the norms of science as a profession. The production of doubt is neither the negative mirror image of the production of positive knowledge, nor simply a scientific Potemkin village where politics is camouflaged as science. Given the extraordinarily high stakes involved, it may be comforting to believe that, while highly damaging to the credibility of science, the systematic production of doubt is also completely external to it. It is politics, not science. I share the agnotologists’ sentiments and commitments, and believe that science and the corporate production of doubt are driven by radically different goals. Sadly, however, they are not easily separable in a formal, conceptual sense. The strategic production of doubt is a very specific form of parasitism that cannot exist outside of its host. Its effectiveness derives only from mobilizing the authority of science against itself.

This presentation asks whether and how technical tools can help facilitate the inspection of images and what steps need to be taken to support tool development. We will discuss some of the factors that speak for and against the use of technology and have a look at the image integrity database (IIDB), which we have build in Berlin to provide training data for algorithm development.

Rapid publication of results — particularly on preprint servers — has grown dramatically during the COVID-19 pandemic, and has forced researchers, health care professionals, journalists, and others to grapple with the concept of reliable and actionable information. The pandemic has given rise to more than 80 retractions at the time of this writing. Is that cause for concern? My lens for this talk will be ten years of experience reporting on retractions for Retraction Watch, including creating the world’s most comprehensive database of retractions, with close to 24,000 and counting.