# Hacking Academia: Data Science and the University

A reflection on our SciFoo breakout session, where we discussed issues of data science within academia.

Almost a year ago, I wrote a post I called the Big Data Brain Drain, lamenting the ways that academia is neglecting the skills of modern data-intensive research, and in doing so is driving away many of the men and women who are perhaps best equipped to enable progress in these fields. This seemed to strike a chord with a wide range of people, and has led me to some incredible opportunities for conversation and collaboration on the subject. One of those conversations took place at the recent SciFoo conference, and this article is my way of recording some reflections on that conversation.

SciFoo is an annual gathering of several hundred scientists, writers, and thinkers sponsored by Digital Science, Nature, O'Reilly Media & Google. SciFoo brings together an incredibly eclectic group of people: I met philosophers, futurists, alien hunters, quantum physicists, mammoth cloners, magazine editors, science funders, astrophysicists, musicians, mycologists, mesmerists, and many many more: the list could go on and on. The conference is about as unstructured as it can be: the organizers simply provide food, drink, and a venue for conversation, and attendees put together breakout discussions on nearly any imaginable topic. If you ever get the chance to go, my advice is to drop everything else and attend. It was one of the most quirky and intellectually stimulating weekends I've ever spent.

The SciFoo meeting is by invitation only, and given the incredible work of other attendees, I'm still not quite sure how I ended up on the invite list (it was perhaps the worst flare-up of impostor syndrome I've ever had!) I forced myself to get over it, though, and teamed-up with Chris Mentzel, a program director in the Moore Foundation, and led a session: we called it Hacking Academia from Inside and Out. The session was in many ways a conversation around the general topic of my Brain Drain post, though it was clear that many of the folks in attendance had been thinking in these terms long before I penned that particular essay.

## The Problem

The problem we discussed is laid out in some detail in my Brain Drain post, but a quick summary is this: scientific research in many disciplines is becoming more and more dependent on the careful analysis of large datasets. This analysis requires a skill-set as broad as it is deep: scientists must be experts not only in their own domain, but in statistics, computing, algorithm building, and software design as well. Many researchers are working hard to attain these skills; the problem is that academia's reward structure is not well-poised to reward the value of this type of work. In short, time spent developing high-quality reusable software tools translates to less time writing and publishing, which under the current system translates to little hope for academic career advancement.

In my Brain Drain post, I observed the rise of data-intensive research and lamented that researchers with the requisite interdisciplinary knowledge are largely unrecognized and unrewarded in academia, even as they are highly-sought-after in the tech industry. I previously labeled this type of person a "new breed of scientist", but since then it's become clear that the working label for this type of person has become (for better or worse) a data scientist.

## Defining Data Science

The term "Data Science" generally seems to get a bad rap: it's variously dismissed as misleading, an empty buzzword, or begrudgingly conceded to be flawed, but useful. Perhaps "Data Scientist" can be understood as just a more subdued term for the "sexy statistician" that Hal Varian predicted would become the top career of this decade.

I think the best illustration of data science's definition comes from Drew Conway's Data Science Venn Diagram, which applies the label "Data Science" to the intersection of hacking skills, statistical knowledge, and domain expertise.

The key is that in addition to the normal depth of knowledge in one's own field, there as a rare breadth to the knowledge and skill-set of a data scientist.

In the words of Alex Szalay, these sorts of researchers must be "Pi-shaped" as opposed to the more traditional "T-shaped" researcher. In Szalay's view, a classic PhD program generates T-shaped researchers: scientists with wide-but-shallow general knowledge, but deep skill and expertise in one particular area. The new breed of scientific researchers, the data scientists, must be Pi-shaped: that is, they maintain the same wide breadth, but push deeper both in their own subject area and in the statistical or computational methods that help drive modern research:

Perhaps neither of these labels or descriptions is quite right. Another school of thought on data science is Jim Gray's idea of the "Fourth Paradigm" of scientific discovery: First came the observational insights of empirical science; second were the mathematically-driven insights of theoretical science; third were the simulation-driven insights of computational science. The fourth paradigm involves primarily data-driven insights of modern scientific research. Perhaps just as the scientific method morphed and grew through each of the previous paradigmatic transitions, so should the scientific method across all disciplines be modified again for this new data-driven realm of knowledge.

Regardless of what metaphor, definition, or label you apply to this class of researcher, it is clear that their skill set is highly valuable in both academia and industry: the brain drain that many have observed comes from the unfortunate fact that academia fails to properly reward the valuable skill set of the data scientist.

## Our Discussion: Academia and Data Science

With this label in mind, our SciFoo discussion focused largely around the following questions:

1. Where does Data Science fit within the current structure of the university?
2. What is it that academic data scientists want from their career? How can academia offer that?
3. What drivers might shift academia toward recognizing & rewarding data scientists in domain fields?
4. Recognizing that graduates will go on to work in both academia and industry, how do we best prepare them for success in both worlds?

I'll go through some of the thoughts we discussed below:

### 1. Where does Data Science fit within the current structure of the university?

The question of data science's place in academia drew a variety of responses and ideas:

The "Fourth Paradigm": data science is simply a label for a new skill-set, and shouldn't be separated from the departments in which it is useful. The thinking here is that data science is simply an umbrella term for an essential skill in modern scientific research. For example, laboratory biologists are dependent on pipetting skills: this doesn't mean that the university should create a new "Department of Applied Pipetting". On the contrary, it simply means that pipetting technique should be part of a laboratory biologist's normal training. Similarly, departments across the university should simply incorporate relevant data science techniques into their normal curriculum.

Data science as a consulting service. Perhaps data science is more like Information Technologies (IT). All modern science labs depend on some sort of computer infrastructure, and most universities long ago realized that it's counter-productive to expect their specialized researchers to effectively manage that infrastructure. Instead, IT organizations were created which provide these services to multiple departments. Data science might be treated the same way: we can't expect every scientist to be fluent in the statistical and computational methods required to work with large datasets, so we might instead out-source these tasks to data science experts.

Data science as an applied computer science department. There has been a trend in the 20th-century of academic subjects splitting into "pure" and "applied" sub-domains. Many Universities have departments of "applied math" and "applied physics", which (loosely speaking) distinguish themselves from the non-applied version by employing the techniques of the field within practical rather than theoretical contexts. Perhaps data science is best viewed as an applied branch of computer science or of statistics which should become its own academic department.

Data science as a new role for libraries. It is no secret that digitization is changing the role of libraries on university campuses. The general public thinks of libraries little more than warehouses for books, but those in the field see printed books as just one particular manifestation of their focus, which has always been data curation. Many library scientists I've talked with recently are excited about the role that new methods and technologies can play in this task of curating and extracting information from their stores of data. From this perspective, Library & Information Science departments may be a natural home for interdisciplinary data science.

Data science as a new interdisciplinary institute. A middle ground to the above approaches may be to organize data science within an interdisciplinary institute; this is a common approach for topics that are inherently multi-disciplinary. Such institutes are already common in academia: for example, the University of Washington is home to the Joint Institute for the Study of Atmosphere and Ocean, which brings together dozens of department, schools, and labs to collaborate on topics related to the climate and the environment. Perhaps such an umbrella institute is the place for data science in the University.

### 2. What is it that academic data scientists want from their job? How can academia offer that?

Moving from university-level issues to personal-level issues, we brainstormed a list of goals that drive data scientists within academia and industry. While scientists are by no means a homogeneous group, most are driven by some combination of the following concerns:

• Salary & other financial compensation
• Stability: the desire to live in one place rather than move every few years
• Opportunity for Advancement
• Respect of Peers
• Opportunity to work on open source software projects
• Opportunity to travel & attend conferences
• Flexibility to work on interesting projects
• Opportunity to publish / freedom from the burden of publishing
• Opportunity to teach / freedom from the burden of teaching
• Opportunity to mentor students / freedom from the burden of mentoring students

I've tried to generally order these from "perks of industry" to "perks of academia" from the perspective of a recently-minted PhD, but this rough one-dimensional categorization misses the variety of opportunities in both worlds. For example, some academic research scientists do not spend time teaching or mentoring students, and some tech industry jobs contain the type of flexibility usually associated with academic research.

Those younger participants in our conversation who have most recently been "in the game", so to speak, especially noted problems in academia with the first three points. Compared to an industry data scientist position, an academic post-doc has some distinct disadvantages:

### Michigan State: new Applied Computation Department

Another active participant in the discussion was Steve Hsu, Vice President for Research at Michigan State University. He shared some exciting news about a new development at Michigan State: the creation of a new department, tentatively called Applied Computation and Mathematics. According to Steve, MSU will soon be hiring around twenty interdisciplinary tenure-track faculty, who will have one foot in their home department and one foot in this new department. MSU already has many impressive researchers working in data-intensive areas across the university: with this effort I'm excited to see the dynamic interdisciplinary community they will build.

## Conclusion

The SciFoo discussion was excellent, as was the weekend as a whole: I would like to take this chance to thank O'Reilly Media, Digital Science, Nature Publishing Group, and Google for making it all possible.

Nearly a year after my Brain Drain brain dump, I am thrilled to see that there are so many good people thinking about and working on these problems, and so many real solutions both in process and on the horizon. While the lure of a well-funded tech industry will doubtless still attract a steady stream of scientists away from their academic research, I'm heartened to see such focused effort toward carving-out niches within academia for those who have so much to contribute.