Technology & Truths: First report on #KS4Austin

The fourth Kavli Symposium on Science Journalism was held in Austin, TX on 19-21 February tackling four themes. Maxime Sainte-Marie, a Postdoctoral Fellow at the Canada Research Chair on the transformations of scholarly communication at Université de Montréal, was one of the 52 participants in the event. Mr. Sainte-Marie participated in the symposium thanks to the support of the Fonds de Recherche du Québec. He shares his experiences on #KS4Austin in the article below. 

In accordance with the Kavli Foundation Meetings program’s objectives, the 4th Kavli Symposium on Science Journalism aimed to foster new research initiatives through cross-disciplinary exchanges. Focusing more specifically on new information technologies, this fourth edition aimed at gaining insights from Artificial Intelligence and Data Science experts in order to find ways science journalism could benefit from the most recent advances in these research areas.

Four different themes were discussed at the Symposium: Data Journalism, Protecting Data Sources & Personal Data, Fact Checking & Fighting Misinformation, and AI Implications in Science Journalism. For each theme, presentations were made by specialists in the field, and breakout sessions involving selected participants were held on the last day.

Regarding the Data Journalism theme, Steve Doig and Jennifer LaFleur, respectively professors of Journalism at Arizona State and American University, shared their journalistic experience in dealing with various data collection, storage, and analysis issues. In the related breakout session, special emphasis was put on the importance of properly informing science journalists about the legal framework and implications of a data-related investigation. To that end, the World Federation of Science Journalism (WFSJ) could create a wiki site, aiming at giving relevant legal guidelines to science journalists of as many different countries as possible.

In the Fact Checking & Fighting Misinformation panel, Kathleen Hall Jamieson of the University of Pennsylvania shared interesting evidence and advice concerning reporting strategies aimed at optimizing communication of scientific material to the general public. Amongst the most important and talked about points of this panel was the importance of giving the audience the feeling they are being treated with respect and giving them the opportunity to engage with the data. On this matter, an interesting point was discussed during the question period and in the related breakout session: rather than telling the whole story and offering readers readymade analyses and conclusions, scientific journalism could rely on more interactive data visualization interfaces such as those offered by the Shiny package from R Studio. By allowing readers to browse their way through the results and play with parametrization settings as they please, such data interaction techniques would let them play a more active role in the process, which could also help inhibit ideological biases.

For the Data Protection theme, Simson L. Garfinkel, Senior Computer Scientist at the US Census Bureau, gave an overview of issues and mechanisms relevant to the practice of scientific journalism. He also gave a short presentation on differential privacy, a recent data protection strategy based on the introduction of statistical noise in record entries. Still, at the running-in stage, this project ultimately aims at preventing the recovery of personal information and identification of specific individuals through dataset pairing. Amongst the different ideas put forward at the related breakout session, it was suggested that the WFSJ develop a secure online platform, which would allow datasets to be uploaded and stored for future consultation without any risks of hacking, data loss, confidentiality breach or data alteration. While such platform could also provide data sharing opportunities, many questions were raised about the legality of such a project, which once again shows the importance of informing scientific journalists properly as regards to the different legal frameworks regulating the collection, storage, and use of datasets.

Finally, for the AI Implications in Science Journalism panel, Sam Han, Direction of Data Science & AI at the Washington Post presented Heliograf, Virality, PostPulse, and ModBot, different tools designed by the journal’s IT team to improve customer engagement, newsroom productivity, and advertisement capability.

In the breakout sessions, many expressed the need to open access to these tools, or else develop similar applications through hackathons or data science and AI competitions such as those hosted by DrivenData.

Blog article by Maxime Sainte-Marie, Postdoctoral Fellow at the Canada Research Chair on the transformations of scholarly communication, Université de Montréal.
Montreal, 26 February 2018