Dr. Jeannette Wing, Corporate VP of Microsoft Research, gave a “Visionaries in Technology” Seminar at the Thayer School of Engineering on October 21. Her talk discussed the development of computational thinking in the 21st century and its role across multiple disciplines, as well as the future of computer science as a crucial part of K-12 education.
Dr. Wing described computational thinking as the ability to think of problems in terms of a series of known and unknown variables, and to solve them systematically as if approaching a problem in computer science. She claimed that the versatility of software is what distinguishes computer science from all other sciences, but cautioned that computational thinking goes beyond simply knowing how to program. Instead, it focuses more on how to apply that knowledge across different disciplines, stating that the proper use of software allows scientists to defy the laws of nature, and create models or test hypotheses without being constrained by the physics of reality. In addition, because computational thinking relies on universal concepts—for example, understanding how to write and use algorithms, or being able to model trends—it is useful in every field, not just scientific research. Applications range from streamlining coffee stations in workplaces, to identifying the genetic basis for stem cell pluripotency, to helping banks identify credit card fraud, to helping people overcome language barriers. Dr. Wing announced that just last week, Microsoft’s voice recognition software reached near human capabilities in speech recognition, which will have significant implications for Microsoft’s digital assistant “Cortana,” as well as in their real-time translation software for Skype.
Computational methods also played a part in the political scene in the 2008 presidential elections. Most polls, such as Gallup or Pollster.com, that are attempting to gauge the zeitgeist of the nation seek representative samples to conduct surveys. That is, these polls try to randomize their sample by picking people from all different backgrounds and socioeconomic classes. However, in an experiment conducted by Microsoft researchers, machine learning was used to understand non-representative data, and the results were more accurate than representative sampling. In their study, the researchers collected demographic data and political opinions from Xbox users, which constituted a non-representative sample because certain demographics are more likely to use Xboxes. The data was then classified into categories (e.g. white males aged 18-25 from California), and used to train a program that calculated the probability that an individual would vote a certain way, given their demographics. The outcome predicted by the program was closer to the results of the election than any existing poll, suggesting that machine learning could be applied to understand human behavior without the need for representative data from a random sample.
Dr. Wing also mentioned a few of the challenges that the US is currently facing in adopting a national computer science curriculum for K-12 education. While the UK has begun offering computer science classes to all 7th graders using a programmable piece of hardware known as the Microbit, the US educational system has been slow to adopt a national standard for computer science education. This is in part due to differing opinions as to how to best teach computational thinking to K-12 students. Furthermore, the US has a much larger population, making it more difficult to enforce a national standard. There is also debate within colleges as to how computational research can be integrated into different departments. However, as computer science and computational thinking play an increasingly larger role in society, there is no doubt that the next generation will play an active role in contributing to and shaping this technological revolution in the coming years.