(Introduction from 2015) I've been thinking quite a bit recently about the question of impact. Like all academics, I hope to have some kind of lasting intellectual impact, but for me a key goal is to find ways that the things I can do might have a more immediate social impact, as well. I had a conversation about this recently that resulted in the précis below, and I thought I'd add it to my Web page.
(Update, September 2017) Two things I'm really thrilled about. One is the response I've been getting personally in response to this page, especially from prospective students, who have been letting me know they care about impact, too. Yes, for many people on the "methods" side of computational linguistics, the interest is in improving the technical state of the art and the specific application is secondary as long as it's interesting and produces publishable results. But more and more I've been hearing from people who would love to connect their technical prowess to work that can affect the world in a positive way.
The second thing I'm really happy about is the way the broader community has been thinking more and more about research impact, something which existed before but feels like it has picked up over the past several years. I'm seeing more critical mass as people look at applications of data science for social good, and I'm also seeing increased attention to the complex ethical questions that our work can give rise to.
Here are some updated thoughts on particular areas where impact is something I care about a lot.
- Language data and mental health. The facts and figures regarding the mental health crisis in this country are staggering; to cite just a few numbers, between 1996 and 2011, annual expenditures on mental disorders rose from $35.2B to $113B, some 25 million American adults will have an episode of major depression this year, suicide is the third leading cause of death for people between 10 and 24 years old, and 89.3 million Americans live in federally-designated Mental Health Professional Shortage Areas. There is an extremely promising line of research that is on the upswing that may help address these issues: using people's language, e.g. what they say on social media, as a source of evidence for early detection and/or monitoring. Recognizing this as an area where the R&D community was starting to see serious activity, I (working with Rebecca Resnik, a clinical psychologist, and Microsoft researcher Meg Mitchell) instigated the first-ever Workshop on Computational Linguistics and Clinical Psychology in Baltimore in 2014. That has since become an annual event, thanks to the energetic organizational skills of others, notably my former postdoc Kristy Hollingshead. I am currently working on a collaborative project with Prof. Deanna Kelly of UMD's medical school looking at social media language analysis in connection with schizophrenia and depression, and I am also collaborating with colleagues and students looking at ways to identify risk of suicide based on social media postings.
- Computational social science. In their 2009 Science article, “Computational Social Science”, David Lazer and colleagues wrote that our online activity “leaves digital traces that can be compiled into comprehensive pictures of both individual and group behavior, with the potential to transform our understanding of our lives, organizations, and societies.” This is an area of great interest to me, particularly because many of those digital traces involve language behavior (social media being the most obvious example) and much of our individual and group behavior is strongly influenced by the language we consume, e.g. political and media framing. There's been a great deal of recent attention on how people make decisions under the general umbrella of "behavioral economics", but far less attention on the connections between language and decision making -- however, Congressional hostility to social science research (e.g. http://www.washingtonpost.com/blogs/monkey-cage/wp/2014/03/13/house-bill-would-cut-social-science-funding-by-42-percent/) threatens to undermine social science research just as it's building momentum toward evidence based approaches driven by large-scale data. The situation is complicated further by private companies stepping into the research vacuum; I organized a panel to draw attention to the issues at South by Southwest Interactive (SXSW) in 2015. I've also begun engaging quite a bit with the public opinion research community, particularly at the American Association for Public Opinion Research (AAPOR) conference, where in May 2017 I co-taught a course on text analysis with qualitative research guru Andrew Stavisky, An Introduction to Practical Text Analytics for Qualitative Research.
- Academia and entrepreneurship. While in academia I've also been an entrepreneur, with experience that includes being technical co-founder of CodeRyte (clinical natural language processing, acquired in 2012 by 3M), lead scientist for Converseon (spearheading development of their sentiment analysis platform), an advisor to FiscalNote (tracking, analysis, and forecasting of legislative and regulatory information), and founder of React Labs, which commercialized my research on scalable real-time response measurement and engagement using mobile devices; in addition, I'm active in consulting. As a result, the interaction between academia and entrepreneurship is something I'm quite interested in, not least because my own primary area of academic research, natural language processing, has recently been identified as an up-and-coming industry that will be worth $13.4B by 2020 with a growth rate of more than 18% over the next five years (http://www.prnewswire.com/news-releases/natural-language-processing-market-worth-134-billion-by-2020-507411291.html)! I enjoy helping to guide people from industry in the right directions with respect to language technology in order to solve real world problems, and helping students and others in academia understand and navigate the landscape where the startup world and industry are concerned.
- Electronic health records. There are two huge issues here that require attention. The first is that the HIPAA privacy regulations were written without the needs of research in mind, and as a result, large scale, data-driven research is fully a decade behind the state of the art in healthcare as compared to other areas. It would literally be easier for me to get a top-secret security clearance than it would be for me, in my academic hat, to get access to a really large set of clinical records from a large and diverse set of hospitals or health care providers. Second, there has been a strong tendency in the electronic records world to force doctors in the direction of click-the-boxes structured input, to the detriment of language data and the clinical narrative, and this threatens to compromise both research and quality of care. I have been up on the soapbox about both issues for quite some time -- for example, see http://schedule.sxsw.com/2012/events/event_IAP10361, http://www.healthleadersmedia.com/page-1/TEC-277956/Are-EMRs-Killing-the-Clinical-Narrative. On the first issue, I've frankly given up on the idea that HIPAA can be revisited, but I have actively been exploring alternative solutions based on data donation, including a collaboration with Qntfy (pronounced "Quantify") to collect data for my project on language use and schizophrenia.
Back to Philip Resnik's home page.