M+E Connections

By Arun Varadarajan, Global Practice Leader of Information Management Solutions, Cognizant Digital Business –

As we dive headlong into a world dominated by artificial intelligence (AI), questions on the readiness of businesses and society to adopt machine thinking remain unanswered. With the advent of big data technologies, organizations have better access to a wide variety and volume of data than ever before.

The key question is whether they have the data they need to solve real-world problems.

Let me illustrate what I mean. My car GPS tracks my location, it notices that I am “moving” at zero miles per hour and that I had 12 sudden stops in the last 30 minutes. That is just one part of the story. Sometimes on Saturdays, I go off the beaten track, taking a detour through one of the trails. I “move” at zero miles per hour — to let a deer pass or to move aside a fallen branch.

I still make 12 stops in 30 minutes — but this time — strictly to enjoy the view. There is another angle: my son just turned 16. He takes out my car at night to impress his lady friends. He sometimes accelerates for no reason, makes several abrupt stops until he gets out of our neighborhood, reaches his girlfriend’s house, and then the data gets all murky. All we know is that the car “moves” at zero mph while he is hanging out.

What does the car company know about me? That I drive the same routes; I am not a top speed kind of person; and traffic is my biggest driving frustration. Do they know enough about me from the data they have acquired from my GPS? Do they know which car my son will buy or if will I stay with the same car when I decide to retire?

Big data technologies nicely capture real-world data, but due to a lack of context (i.e., situational awareness) they occasionally fail to provide real insights. Why is context important? Well, some of the biggest problems the businesses want to solve are based on human behaviors – and our understanding of them.

What if insights begin with humanism, empathizing with and designing things for real people? What if things were designed to place human experience at the center of the art-of-possible?

Once these deep-rooted, nuanced, cultural and historical contexts are well understood, techniques such as big-data and AI/ML (machine learning) algorithms can be used to observe and respond to human needs.

Applying social scrutiny

By scrutinizing design thinking and using social science, businesses can get a deeper understanding of how to model the world better in this context. Social scrutiny is accomplished by conducting deep immersive psychographic and ethnographic research where anthropological techniques are applied to study human behavior in a given circumstance.

In my view, often non-contextualized data sets are potentially being used to train deep learning solutions that will over time orchestrate a good part of our world. This could lead to unwanted and dangerous outcomes. As AI tools are applied to inform decisions that are based on human behavior, the lack of data contextualization will be amplified, delivering failed outcomes that ultimately impact human experience.

For example:

* With the rise of fintech innovators and e-commerce companies, a completely new digital shopping ecosystem has emerged that includes organized rings that perpetrate fraud. Despite real-time monitoring systems that run sophisticated algorithms and deep learning models, fraudsters always seem one step ahead of commercial enterprise.

To understand what makes fraudsters tick, we assembled some of the world’s best anthropologists and ethnographic researchers to study the behaviors and psychology of fraud. We did this by spending time in jail with fraudsters, speaking with victimized merchants and engaging NYPD officers. Repeatable fraud is typically perpetrated through elaborate rings. An online retail fraud ring starts with hackers who have unprecedented quantities of stolen card information.

Other key players are, of course, the fraudsters and the buyers who develop trusted relationships, so deep that fraudsters plan their heists based on buyer preferences. Such operations involve a new login, new drop location, new merchant and a new card, which spawns a rapid spending spree. Understanding how these fraud rings operate and how they behave was key to improving our algorithmic models and AI techniques. This enabled us to create meaningful context for the data that informed these models.

* Another case that is even closer to human behavior is that of a racing theme park business – where management had struggled to design the right guest experience and personalization model. The team eventually decided that before making big commitments it should conduct immersive research, based on an open inquiry modality, that examined the social characteristics and cultural contexts of the people likely to visit their theme parks.

The study found that individuals interested in racing and cars do not visit such theme parks. Families do. Families, in fact, with at most a single racing enthusiast. Families who seek to create memories together. Mostly families who have traveled and seen enough to have a keen eye for authenticity. Using this human-centric lens, the park then determined the right instrumentation and guest measurement strategy to feed their models. This, in my, opinion will keep the park relevant and ensure that it continues to meet the expectations of its guests.

In conclusion, consider the following:

1. What if we juxtapose insights through social scrutiny with traditional data?
2. What if deep learning models are enabled for proactively seeking and capturing behavioral data?
3. What if we include social scientists and ethnographers in AI teams?

There is much to ponder and much to debate, and this is just one key element that businesses must examine in greater detail before AI runs amok.

—-

Click here to translate this article
Click here to download the complete .PDF version of this article
Click here to download the entire Spring/Summer 2018 M&E Journal