How Is Behavioral Science Related to Data Science?
Posted on Feb 06, 2020
In this article we explore the endless debate between the world of behavioral science and data science. Though both behavioral and data science share many common properties, how one is trained to think and approach problems may differ markedly between the two related disciplines. At the end of the day, what adds value to your firm depends on whether you are analyzing and manipulating internal databases or expanding your knowledge of customers with new data.
How exactly is behavioral science different from data science?
It is helpful to begin by taking a look at what makes behavioral science unique from data science. According to Investopedia, data science can be defined: "Data science provides meaningful information based on large amounts of complex data or big data. Data science, or data-driven science, combines different fields of work in statistics and computation to interpret data for decision-making purposes." Conversely, behavioral scientists conduct experiments and analyze data in light of scientific findings on what drives humans to make the choices and decisions that they do. This leads to the first difference between behavioral science and data science proper.
(1) Behavioral scientists possess deep knowledge of psychological and economic theory to make applied and theoretical advances in the study of human behavior.
Over years of training, a behavioral scientist comes equipped with a long list of scientific studies, findings, and outcomes from rigorous research agendas. They can also tell you the limits of science when applied to human behavior, as many problems in social science require "stylized" facts or working theories. You want a behavioral scientist when you need to go beyond existing data. Without expertise on what drives human behavior, you will end up with "folk theories," which is code for anything someone may have come up with, whether that's someone you know, in our society, or inherited "wisdom" as temporary explanations or mere gossip ("old wive's tales").
Behavioral science and data science share many common properties, but the day to day of each job title varies. When one calls him/herself a behavioral scientist, it implies first and foremost a foundational understanding of human behavior along with multiple methods to explore, explain, and predict human behavior.
(2) Data collection is what behavioral scientists do best.
Data science is the computational study of data, how to manipulate, merge, combine, model, and make predictions. In the sense that both fields are analyzing data to advance academic and industry insights with more useful models and data flows, they share a lot in common. Yet, to call oneself a behavioral scientist usually means an emphasis on assessing data variables and collecting new data in a systematic and/or scientific manner to predict behavior.
Causal prediction is at the heart of both data science and behavioral science. It is more difficult and possibly inaccurate to make historical data speak up with predictive models than simply creating a new study and collecting data in a more systematic manner. For this reason, behavioral scientists often possess expertise in economic psychology, social psychology, or economics. Data scientists can learn their trade relatively quickly because the focus is on managing data than creating it. For example, there are 3-month bootcamps in which one can become a data scientist with popular programming and database management languages such as Python, C-sharp, SQL, and R.
For the most part data scientists manipulate and model historical data from companies, such as customer purchases, volumes, and transactions. Behavioral scientists will identify new data vectors and ways to make sure that quality data is collected and maintained. They will be able to show you how some data types are hiding key contextual details and so may mislead you when you look at multiple demographic groups, locations, and purchasing environments.
(3) Behavioral scientists nuance data variables to map them onto real problems by integrating perceptions, beliefs, opinions, preferences, and choices.
Probably what sets behavioral science apart from data science the most is the use of multiple data types together to map them onto real problems. That means that a behavioral scientist wants to know the decisions made, the beliefs of the decision maker, the social effects and context, and the outcomes. If one analyzes only transactional historical data, these details are lost. A person's decisions are treated as proxies for their beliefs and preferences. A customer's repeat purchases are treated as substitutes for their loyalty and trust in the product. As behavioral scientists could show, there could be a range of reasons that drive people to choose within context, such as time scarcity or misinformation. Similarly, purchase behavior may be the result of circumstances, such as inertia or inability to find a new phone plan, for instance. These will lead companies who hold the data to believe that customers are simply that: customers. And thus expect their return. A data scientist could tell you that your customers bought 3.5 socks on average, but a behavioral scientist could tell you more about the why and the when, to see if those trends will continue.
(4) The word 'Experiments' means something different to behavioral scientists than it does to data scientists.
For behavioral scientists, data collection is at the heart of everything. To run a randomized or quasi-randomized experiment, there must be a way to track and collect data on decisions in real-time whether about decision-making itself (behavioral decision science) or actions that result from decisions (behavioral research), or even the purchasing patterns that result from behavioral drivers (as is the case in behavioral marketing). A behavioral scientist would set up a study, put customers into treatments and control groups, then identify which of the treatments has the most uptake of a product or service. Data scientists run experiments by simulating data similar to an econometrician. They can find "random" variables (things that are 'as good as random' in the world), but they do not introduce true randomization, because they cannot setup the experiment from the outset.
What seems clear is that when behavioral "x" likely comes up, it involves similar data manipulation, cleaning, structuring, and visualizing that many data scientists are also experts at.
There are many ways to frame the timeless debate between the data curators and the behavioral enthusiasts. What seems clear is that when behavioral "x" likely comes up, it involves similar data manipulation, cleaning, structuring, and visualizing that many data scientists are also experts at. On the one hand, behavioral scientists believe that they are more engaged in scientific discovery, taking theories and applying them to the real-world in industries, such as health, finance, pharmaceuticals, and insurance. For this reason, they tend to view behavioral science as distinct from data science. Data scientists, on the other hand, have often had a more agnostic stance. In our professional opinion, behavioral and data science are partners in what industry calls culling, curating, structuring, tracking, and collecting data.
Many stakeholders and leaders at companies do not know what behavioral science is, and they barely understand the idea of data science. More needs to be done to make a bipartisan effort to instruct non-data enthusiasts to collect, maintain, and manipulate data for the greater good of our decision-making.
Let's agree to disagree with anyone who thinks data does not matter. At the end of the day, for both behavioral scientists and data scientists..
Data has a better idea.
© Behiring 2020