Insights were shared on how Data Science informs evidence-based policy decisions and how it can truly establish transdisciplinary approaches. A Data Science Ethos tool was presented to help data science teams embed ethics and societal considerations at each stage of the data science workflow.

Presentation 1: Decoding Climate Vulnerability: Harnessing the Power of Data Science and Causal Inference 

In the first presentation, Prof. Francesca Dominici convincingly demonstrated why data science is necessary to make evidence-based decisions. Fighting climate change and air pollution is a good example of where data science can make a positive impact on the world.  

Because of data science, in 2017 it became clear that even safe levels of air pollution in the United States were deadly. Publication about this evidence was met with a lot of push-back, and the question of causality of air pollution and mortality rates was raised. Data science offered a constructive response to these criticisms; data science techniques such as machine learning and casual interference disentangled the causes from confounding variables, and enabled better identification of significant differences between individuals as a direct result of air pollution. Machine learning and data science also allowed for better measurements of air pollution exposure and pinpointing of susceptibility and vulnerability.  

Thanks to insights from data science, the steps that are needed to mitigate climate change in future are similar to the steps needed to reduce air pollution in the present, such as cutting back on burning fossil fuels and biomass. These findings demonstrate that data science has the potential to change public health policy guidelines and laws in relation to safe levels of air pollution and also future climate change laws in the United States.  

Presentation 2: Advancing on Research Assessment: the role of research data in a new evaluation landscape 

Prof. Eva Méndez started the second presentation with the observation that although data science has fundamentally changed the way we do science, this is not visible in the way we measure research outputs. The centrality of the Journal Impact Factor in research evaluation is the most important bottleneck of meeting commitments to open science - science that is for everyone and can actually make an impact. 

The University of Amsterdam, alongside many other Dutch universities, has signed an Agreement committing to Reform Research Assessment (ARRA). We committed to recognize the diversity of contributions and avoid the use of rankings of research organizations, but this is not happening yet in practice. To bring about change, we need more seductive ways to measure the outputs of science.   

Data science can be the connecting dot between different disciplines, and this is where Data Science Centres can make a difference. Understanding the value of data and software by considering it as a first class research output can be the key to meeting these commitments. Data Science Centres can bring together diverse and interdisciplinary groups of people, and support greater sharing of data between researchers.   

Presentation 3: The Data Science Ethos: Operationalizing ethics in data science

The third presentation was by Dr. Micaela Parker, director of the Academic Data Science Alliance (ADSA). In the standard data science lifecycle, nowhere can you find the term ‘ethics’ or ‘societal impact’. That is why ADSA launched the Data Science Ethos tool. 

The Data Science Ethos tool departs from and represents a more true-to-life model of data science workflow that embeds ethics. At each step of the cycle, it shows how ethics can be integrated by suggesting questions that can be considered.  

Data science is a constantly evolving field, and by bringing together data science community and data science leaders you can set the direction. ADSA is a community building initiative supporting university researchers and educators to learn, use, and teach data-intensive methodologies and responsible applications. ADSA aims to connect, engage, and empower data science researchers and educators from different disciplines to discover, build, and share resources and opportunities.  

Panel discussion 

The morning concluded with a lively and interactive panel discussion with the audience. A variety of topics and questions were discussed. 

Ethical data science

Answering a question from the audience on ethical awareness in data science, all speakers stressed the importance of bringing data literacy into the classrooms. Francesca Dominici: “Ethical data science needs to be embedded, not only in data science life cycle, but in the human life cycle.” Concerns were raised on the fact that much of the data science driving everyday technologies will remain a black box. Micaela Parker shared her take on the huge responsibility of people who develop algorithms: “We drive cars, even though we do not always have the knowledge how the car exactly works. We must be able to trust the people who have built them.” Dominici pointed out that we are not there yet. “Bias remains an important issue. The worry is that technology moves at a pace that is 10 times faster than responsible use.”  

Inclusive approach

And what is data? Numbers, words? What about pictures, silences, life stories...? Answering the question raised by an anthropologist in the audience, the speakers confirmed that data can be anything. This means we also constantly need to revisit (ethical) codes. There is no one size fits all. Elaborating on the question of multidisciplinarity, another person in the audience wondered if we weren’t glorifying data science. What about more qualitative approaches? Eva Méndez emphasised that we need an inclusive approach to have a good view of the world. Learning social aspects, and qualitative interviewing that goes into researching any problem, is important: we need both.  

Academia and industry

Another question from the audience was how data science in universities relates to industry. Francesca Dominici shared her opinion that academia cannot advance in data science if they do not partner with industry. It is important to consider how we cooperate in a way that is mutually beneficial but doesn’t influence research. The reality is that industry is ahead of us in many types of technologies. Eva Méndez: “The problem is that in the current system, the transfer to the industry comes if I discovered something that I can sell to you. We need to flip the knowledge transfer model.” 

Final takeaways

Moderator Tessa Blanken invited the speakers to share some final takeaways. These very much endorsed the mission of the UvA Data Science Centre to enhance and accelerate the university's research by developing, sharing, and promoting data science methods and technologies across faculties. Micaela Parker: “My hope is that data science leads the way of how all science is conducted, breaking down silos of departments, working across disciplines. There needs to be better dialogue. It’s about communication and collaboration. Ethics and responsible behaviour comes with that.” Eva Méndez: “Currently data science is a separate discipline, but in the future, data science will be the common mechanism that connects everything. The word for the future is: responsibility.”

Event recording