Please distribute to others who may be interested... You are hereby invited to a seminar in our twelfth interdisciplinary series on Evolution, Complexity and Cognition <http://ecco.vub.ac.be/?q=node/108> (ECCO 2016-2017) <https://www.researchgate.net/profile/Marjoriikka_Ylisiurua>
Time: Friday May 26, 14h-16h Place: *room * *D.1.07*, building D, VUB ------------------------------ Seeking evolution from social sciences: Scientific abstracts on dynamic processes as machine learning material Marjorikka Ylisiurua Abstract: Increasing number of scientific publications means any average researchers should, in principle, be familiar with a gargantuan pile of past research that is constantly growing rapidly. In medical and biosciences especially, regular meta-analytical literature reviews on a topic have therefore become the norm. These are summaries of past research, that may also reprocess and combine statistical findings from several experiments, thus improving reliability of conclusions. To help in the labour-intensive task, computer scientists have developed tools and semi-automated machine learning methods.* The benefits of machine learning are enabled by open access publishing, and even commercial publishers allowing (limited) access to their abstract databases via public APIs. This has made it tempting for other disciplines to join in the trend. For multidisciplinary field like consumer economics (subfield of economics with links to sociology, psychology, marketing, science & technology studies, culture studies, nutrition science etc.), the amount of potentially relevant abstracts is multiplied, making machine learning an even more tempting method to approach research literature. Instead of a full-sized literature survey, a so-called scoping review is a lighter, more interpretive approach that seems more suitable for "soft" social scientists whose methods do not hinge on sample sizes and p-values. However, different disciplines and sub-disciplinary theories often discuss recognizably similar concepts, but use different terms. The subsequent need for contextual interpretation leads to a more conceptual problem at the heart of cogntiive science and language technology: how can a fuzzy concept such as "a dynamic process", discussed from various angles in different scientific traditions, be mapped between them all? Technically, the question is already tackled by commercial start ups that engage in scientific artificial intelligence**, but their products admittedly work better with "harder" scientific fields where studied concepts are more rigid. In the higher level of language and cognition, Google Neural Machine Translation is boldly leading the way. So far, the results have sadly not been uniquely impressive***. For my second PhD-article, I've prepared a python machine learning program to scrape Elsevier Scopus API for various disciplines where evolutionary processes are studied. As the actual modeling and interpretation is a work in progress, I will also discuss my empirical experiences on using machine learning in a scoping review, and resulting musings possibly spanning from epistemological notions, to imagined potentials of artificial intelligence that may be capable of independent scientific problem-solving. *e.g. rayyan: https://rayyan.qcri.org/ **e.g. Iris AI: https://techcrunch.com/2016/12/05/iris-is-an-ai-to-help-science-rd/ https://www.wired.com/2016/11/artificial-intelligence-dig-cures-buried-online/ ***https://www.buzzfeed.com/patricksmith/im-a-dick-i-used-to-be-a-hawk ------------------------------ Upcoming Seminars *June 2nd* Vincenzo De Florio Service-oriented communities: A novel architecture for smarter systems and organizations *June 9th* Tomas Veloz Chemical Self-organization Theory and its Application to the Complexity-Stability Problem *June 16th* Forrest Rosenblum TBA See also the ECCO/GBI calendar <https://www.google.com/calendar/embed?src=azMyN252aWluM2JoMnU3MXY5OGt2ZzliOGdAZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbQ> . You can add this calendar to your calendar application through here <https://www.google.com/calendar/ical/k327nviin3bh2u71v98kvg9b8g%40group.calendar.google.com/public/basic.ics> More info about the ECCO seminar program: http://ecco.vub.ac.be/?q=node/108 <http://ecco.vub.ac.be/?q=node/108> -- Cadell ECCO Group (VUB) <http://ecco.vub.ac.be/?q=node/1> Email: [email protected] Website: https://cadelllast.com
