Co-Chairs：Radmila Juric, Sang Suh, Elisabetta Ronchieri

One of the most intriguing buzzwords of the second decade of the 21st century is DATA SCIENCE. It overflows academic curricula all over the world, and appears to be a driving force behind what we perceive today to be artificial intelligence. However, by looking at practices, computational models, software tools and development environments for data scientists, there is hardly any evidence that data science provides a scientific approach to resolving modern computational problems. However convenient it was to place these two words DATA and SCIENCE together, without agreeing on what their joint meaning would be, we are now creating computations, under the umbrella of data science, which diverge from our basic principles of computer science, built over the last 70 years. Problems are numerous. They start with a complete lack of formalism in terms of defining exactly which type of computational model data science generates. It continues with legitimate distorting of semantics of huge data sets, desperately needed for training algorithms, in order to “clean” or prepare/format data to be suitable for pre-prepared algorithms. They assume that noisy data have no semantic; they look at missing data values in data sets as places to be either eliminated or filled with “something”, without knowing their meaning and without having semantic justification for filling them with any data. Tools are allegedly performing data science operations upon your data without telling you exactly how they deal with semantic of data. They obscure algorithms behind them in order to either automate these computations (without your involvement) or perform faster data processing or give answers to any question one may have.

In this workshop we would like to debate important problems of

1. Collecting, formatting and preparation of data for further processing as prescribed in data science
2. understanding the impact of missing data to the overall semantic of the data set and their impact on the performance of the algorithms which use such data set
3. analysing the existence and the role of dark data which has various interpretation in data science and consequently affect results of computations upon them
4. believing that imputations in the data set brings trustworthy solutions in predictive algorithms and as such has to be performed.
5. eliminating noisy data, without understanding its semantics and purpose/role within the data set.

1. Predictive versus logic inference: Would computer scientists agree to using predictive technologies and algorithms and claim the power of predictive inference, without thinking about the manipulation of the meaning of data and using logic with reasoning, in order to claim “real” inference?
2. Is predictive inference a backbone of what is today called AI?
3. Does predictive inference really mean marching forward in modern computing because data science favours it?
4. Is Data Science a culprit which stops us in galloping forward in computer science?
5. Is there anything else Data Science can give us (apart from predictive inference)?
6. Does Data Science eliminate exactly what it should be: a SCIENCE which opens doors to many other innovative views on how to manipulate semantics of modern DATA?
7. Could we agree that predictive technologies are possible companions and not a “sine qua non” of our computations which are expected to pave the way towards artificial intelligence?