SB.S.

Why social science is failing us and what we can do about it

This blog post is a preview of my upcoming book on the current challenges and possible solutions facing the social and behavior sciences (SBS). It is adapted from a response I wrote for a Defense Advanced Research Project Agency (DARPA) Defense Sciences Office (DSO) request for information (RFI) on confidence levels.

As a qualitative researcher and project associate specializing in survey research at the University of Nebraska-Lincoln (UNL) Bureau of Sociological Research (BOSR), I couldn’t help but notice many of these problems outlined in the RFI in our industry.

My suspicion that many SBS research organizations had moved away from the fundamentals of high-quality research were confirmed by the 2016 election polls. Too many are taking methods used in market research, where the results needn’t be as accurate, and applying them to SBS research, where the results absolutely need to be accurate to be useful.

This occurred in large part because SBS researchers do not have access to 21st century tools. Instead, the tools available are either extremely outdated or geared towards market research. Nobody is making precise, state-of-the-art tools specifically aimed at serious, academic researchers in SBS. We could talk ourselves in circles about p-values and confidence levels and how we might tweak complicated models to make things marginally more accurate. But none of this will solve the problem in the long-term and none of it helps non-experts who do not wish to go back to school and earn a degree in statistics in order to make good decisions on the job.

Though the solutions to the problems expressed in the RFI are not currently in existence, my team and I are working hard to create them for the future because we believe humans must be able to solve social problems and make sound policy decisions based on science. Our value system is as follows:

I see the Department of Defense’s (DOD) problem like this; They want to do a good job. They want to make decisions with maximum positive impacts and minimum negative impacts. They want to make rational decisions based on science like they would for anything else. If they’re making decisions about a possible epidemic, they can look at the literature on that. They can find the experts who know the most about it and be confident they’re making decisions based on the most scientifically accurate information humans have.

But when it comes to SBS research, even an expert cannot generally be confident that what they are reading is true. Humans are complicated creatures and until very recently, mainly in the past year, humans have not had the technology necessary to really study and understand ourselves the way we do the physical world around us. We have some rules of thumb through economics, sociology, psychology, etc, but it’s all rather Newtonian. It predicts things based on very specific circumstances and does not help describe general rules that apply to all social creatures.

There are three things that would truly move the needle for the accuracy and usefulness of SBS research to non-experts:

• Utilize recent advancements in cognitive computing to scale the judgements of experts in the field and analyze all literature constantly, making even a non-expert user more up-to-date than any true expert has the ability to be today.
• Utilize recent advancements in cognitive computing and big data analytics to look for patterns in new, qualitative data sources such as weather sensors, local news, social media, blog posts, etc.
• Invest more in basic research for SBS.

Our startup AmeliorMate aims to achieve all three goals simultaneously. We plan to knit together various state of the art APIs, especially from IBM’s Watson, to achieve this. Once a company trains a Waton algorithm, that new algorithm becomes the company’s intellectual property. IBM’s recent work using Watson to conduct literature reviews and provide recommendations for oncologists shows it is possible to create such products.

In addition, Watson is very good at combing through unstructured and big data to find patterns and connections human researchers alone could never identify. For example, because The Weather Company is owned by IBM, our customers will be able to use Watson to look for any connections between the data they have and data collected by IMB’s extensive network of weather sensors.

Basic Research

SBS also suffers heavily from a lack of basic research. Because SBS researchers have not traditionally had access 21st century tools, the tree of knowledge for SBS is a bit Newtonian. Researchers like Alex ‘Sandy’ Pentland at MIT have shown there are some social patterns humans engage in which are just as predictable and can be studied equally well in colonies of bees and ants. For example, in his TEDx Beacon Street talk, he explains how principles of exploration and engagement can predict the health of a human neighborhood as well as the colonies of these smaller social creatures.

By investing in basic research in SBS, we can uncover the most fundamental principles of human interaction, giving a solid foundation for all research conducted thereafter. The problem is, when it comes to our physical universe or even our biological universe, we understand the rules of the game, while in SBS, we don’t. We can engineer rockets that can go to Mars because we have rules to work with. We can create vaccines because we have predictable frameworks in biology as well.

By investing in basic research for SBS, we can learn the rules of our social universe. Only then can we effectively engineer solutions to the massive social upheavals which are part and parcel of the amazing technological advancements being made. I believe our company could do a great deal in helping the DOD solve social problems and prepare for the future by creating a Watson for oncology for people utilizing SBS research to make decisions.

And in doing so, we hope to do our part in creating a world where humans truly do matter most.