Everyday Data Science

You probably don't think of it this way but you're doing data science every day. You're doing machine learning, specifically reinforcement learning, every time you get behind a vehicle's steering wheel.

Before I get into an example, though, a quick aside: Data science isn't easy. I simplify data science by helping organizations get started or improve its function; define the skills and hire the right people; determine infrastructure; pick the right tools; define processes and documentation for consistency and scale; identify and organize or get started collecting the relevant data; unify the function where there can be overlap between teams; coordinate model outputs into existing software or digital assets; determine whether outputs should be real-time or on some time-based replenishment schedule; and explain the outputs from the data science team for the business that make it valuable and actionable.

On to the example…

As you approach an intersection and the traffic light turns yellow, what do you do? You make a decision to go through or stop.

Running the changing light or stopping is a binary decision. There are only two outcomes: run the light or stop. Your brain, and you, make that decision based on the available and historical data at that moment. Let's document some of the factors that impact your decision to stop or go:

• Distance: How far am I from the intersection?
• Pedestrians: Are people or objects in the crosswalk or intersection?
• The weather: will it affect my ability to stop or accelerate in time?
• My passengers: are my kids or someone I'd prefer not to scare with me?
• My speed: can I make it through the light without having to speed up much?
• Traffic: Are the cars ahead of me going through? Are there cars waiting in the opposite left-hand turn lane to finish their turn?
• Consequences: Could I get in trouble if I ran the light?
• Monitors: Are there red light cameras that would record me as I go through if I don't make the yellow?
• Urgency: Am I in a justifiable hurry to get where I'm going?
• Past experience: Have I made the decision to run a light or stop before? What were the consequences?

You probably came up with other factors that impact your decision. The specificity of the data to be applied to the problem is a huge differentiator in defining data science and its impact on decision making. Would you rely on a survey of 10 people in another city that noted 70% of them ran yellow lights more often than not to make your decision? I hope not. But many organizations rely on small survey samples and generalization to make some of their biggest business decisions.

This is what differentiates data science as a tool for making smarter decisions that optimize your chances of success. In the driving case above, your past history of success or failure to run yellow lights reinforces your decision to do it again in the present case. You'll use your past experiences as a guide; that's all reinforcement learning does. As your mom might have told you: learn from your mistakes.

You may have reached the relatively obvious conclusion that the yellow light example is one aspect of a hot topic: self-driving cars. You'd be exactly right.

Like automating anything with technology, data science and machine learning can be negatively disruptive. This is why organizations working to integrate data science effectively understand the implications and consequences of doing it the right way.

If you're interested in talking more about the role of data science reach out to me here or at sam.johnson@bluejacketsol.com. I'm always interested in other perspectives and examples of what works and what doesn't.