IN OUR HEADS
September 28, 2017
Shravya Kaparthi, Associate Director, Decision Sciences, RAPP Dallas
I solved my identity crisis. The solution? Stop worrying about my identity.
Throughout my Media & Advertising career, I have always used numbers to help make decisions. Historically, in media planning, predictive modeling for behavioral segmentation & targeting, campaign analytics/ROI calculations – or even quarterly planning meetings – but, recently, with the increasing conversation around Artificial intelligence, Natural Language processing, Machine learning, and the likes, I began to shy away from calling myself a ‘data scientist’. As a non-computer-sciences business major, I always felt I fell just short of deserving the title. Especially within the marketing realm, where who I am and what I do is more typically called performance analytics, marketing sciences, decision sciences or experience analytics. A plethora of names and ever-evolving roles and responsibilities almost always ensures we worry too much about titles and disciplines.
The data science conference helped reshape my perspective on what matters most – solving problems.
Re-integrating econometrics and machine learning, bootstrapping and bidding algorithms, corporate credit risk analytics and chat bots, deep forecasting and deep learning, online payment fraud detection, and assistive AI for cancer treatment. These are just a few of the focal conversations centered in data science – as both a profession and answer to today’s greatest challenges – at this year’s Data Science Conference in Seattle. We explored its opportunities and limitations, evolving tools and changing skill-sets’ demands and expectations from the community.
Bringing these professions to life, I met medical/health care research professionals, people who work with police departments/forensics using data to detect crimes/identify criminals, e-marketers, financial technologists, airline employees, educators, logistics & operations companies, retail, academia, social networking and search engine leaders. Most interestingly, to me, I met professionals using data to make a difference - helping fight hunger, provide help during natural calamities etc. working on a variety of data-driven solutions to improve the current state of social good. The list goes on and on.
And so, although I’m already steeped in it here at RAPP, I was overwhelmed by the power of this field viewed at such a large scale. I’ve always been aware of its potential, but interacting with passionate data-science professionals helped me appreciate these possibilities tenfold.
Upon beholding the possibilities and realities of my field on such a large and diverse scale at this conference, I realized the name we call ourselves really doesn’t matter. We are curious souls and problems solvers – programmers who get statistics or statisticians dabbling in programming, natural sciences graduates or psychology majors. The exact title and expertise makes no difference. If you use data to make decisions and solve problems or identify new opportunities and optimized solutions, and effectively communicate about the same to non-data folks, you’re a data scientist. But it really doesn’t matter what you call yourself. You are a problem solver.
And here are few key learnings from the conference that will help me become a better problem solver:
1) Power of inferential and counterfactual questions:
Start the process with the right questions.
“An approximate answer to the right problem is worth a good deal more than an exact answer to an approximate problem” – John Tukey
There will be times when the prediction goes in an unexpected direction and our gut and conventional wisdom might contradict what the data shows us. That’s when one should learn to balance exploration and exploitation with further investigation – understand accuracy and interpretability and be smart about conclusions and the way forward.
2) Power of persistence and patience:
No matter how much data was crunched or how many months or years of effort went into a project, small improvements are most vital. Every experiment teaches us something. As long as we’ve learnt something new, it’s worth the effort.
3) Power of context:
People always want to know, “Why? And what are the implications?” But one of the important aspects around these questions is context.
Contextual filters help us bring out the best solutions. Algorithms are directional, not a complete solution, so they must be followed by right judgements, decisions or behavioral changes for the biggest impact. E.g. United Airlines algorithm controversy or IBM Watson and MD. Anderson controversy.
4) Power of design thinking and storytelling:
Human-centric design, powered by logic and common sense, should drive our thinking. Data scientists must think like designers solving problems – not like engineers or statisticians. We should add a grain of psychology for the best solution designs.
Along with employing a specific perspective comes the responsibility of communicating to various stakeholders (most often than not who do not understand data) to influence change. Storytelling helps in creating the right impact, so we should practice the art of embedding numbers in a social/business/consumer context until we’re proficient at it. A proof of concept approach around learnings helps influence change within an enterprise or community.
5) Power of augmented intelligence:
Artificial intelligence (AI) has raised a lot of profound and existential conversations around human vs machine, of which conversations human impact is a central and crucial aspect. No amount of data is a true representation of the entire universe – there will be anomalies, there will be errors – and, as such, one should always use their discretion before making any leaps of faith. Data scientists have to constantly make tough choices – decide parameters, create data filters in the domain-related-context, based on the end goal, etc. These are the choices will help shape the best solutions. Which means, at the end of the day, data science is more art than science.
Most importantly, we shouldn’t be worrying about artificial intelligence, we should be worrying about artificial stupidity! AI should augment human expertise. Technology is the easy part – it’s the social practicing that is complex and ever-evolving! New age of cognitive collaboration between AI and humans should eliminate the machine bias, instead enhancing it with augmented intelligence.
Too often, there is conversation around titles and disciplines, but after the data science conference, I am more committed than ever to learning and investing in skill-sets so that I can be a better problem solver.
