Think About the Human Touch in Simplifying Decisions, Using Big Data and Teaching Machines to Learn

The Human Touch

In this issue of FRAMEWORK, we revisit articles expressing a recurring theme in my thinking:

As we work with and think about data analytical processes, tools and technologies—and machine learning and artificial intelligence applications—we need to always remember the importance of human direction, intelligence and insight.

As Alma Derricks of Cirque du Soleil says: “Big data is not a silver bullet and can’t replace instinct, intuition, guts or empathy.”

It’s our data experts, computer scientists and business professionals who use the tools and technologies, working together to find meaning at the data analytics’ intersection of art and science. Our art comes from our “human touch.”

With Marcia Tal’s insightful point-of-view and commentary, Tal Solutions’ FRAMEWORK newsletter presents you with a bi-monthly collection of curated, personally selected content and media focused on big ideas in and beyond Data Analytics.

FRAMEWORK : Data Analytics

Add a Human Touch to Banking’s Information Business

“More than ever, banking is as much an information business as it is a financial business,” claims Bryan Yurcan in his American Banker article “Why Small Banks Need Big Data.”

Further, he warns: “With the vast amount of data banks have on customers and transactions and spending habits, those that aren’t effectively mining this data risk falling into irrelevance.”

Although compliance, risk management and fraud prevention are popular areas of analytical focus, solidifying the foundations of business processes is where data analytics is critical. I always think about Atul Gawande’s The Checklist Manifesto – as we implement new strategies using our data, do we have true checklists along the way and audits at critical points?

We must ask ourselves: Are we using our data to ensure we are in compliance with the law AND are we providing our customers access to the most relevant products and services for them?

We work at the intersection of art and science. Today, banks have unprecedented access to an array of advanced analytical tools and technologies – the science. We need to blend the application of these tools with the human touch – the art.

As I wrote in a blog series “Why Analytical Modeling Will Never Be The Same,” three human roles are necessary for effectively using predictive analytics and machine learning to achieve financial impact:

  • Data Experts
  • Computer Scientists
  • Business Experts

When I spoke with Henry Zelikovsky, CTO, Director of Engineering at Starpoint Solutions, he succinctly captured the human touch:

“We use technology and machine learning in order to process an enormous amount of data in a predictable manner to give to a person better qualified information within our decision support tool…but the decision is from the person.”

So, when you and your fellow experts unfold your Swiss Army Knife of data, analytics and technology to solve banking challenges and build business solutions, remember that these are tools – and “don’t forget the human touch.”

FRAMEWORK : Data Analytics

Join a Deep Dive into the Intricate Murky Regularity in Data

In reading Gideon Lewis-Kraus’ New York Times Magazine article  “The Great A.I. Awakening, we join an exploration of “how machine learning is poised to reinvent computing itself.”

Lewis-Kraus introduces us to this awakening by relating how a Google Brain team used neural networks in an A.I. machine-learning system to exponentially improve Google Translate—literally overnight. Converted to the A.I. system from a traditional programming platform, Translate “had demonstrated overnight improvements roughly equal to the total gains the old one had accrued over its entire lifetime.”

Google Brain, founded five years ago on “the principle that ‘neural networks’ that acquaint themselves with the world via trial and error, as toddlers do, might in turn develop something like human flexibility.”

As Lewis-Kraus reports, Google’s CEO Sundar Pichai differentiates between current applications of A.I. and the ultimate goal of artificial general intelligence – “Artificial general intelligence will not involve dutiful adherence to explicit instructions, but instead will demonstrate a facility with the implicit, the interpretive.”

Here’s the leap in learning: “If an intelligent machine were able to discern some intricate murky regularity in data about what we have done in the past, it might be able to extrapolate about our subsequent desires, even if we don’t entirely know them ourselves.”

This method of machine learning with pattern recognition is transferable to almost every property and domain. As the article concludes: “Once you’ve built a robust pattern-matching apparatus for one purpose, it can be tweaked in the service of others.”

Citing example after example, we see this transferability:

  • A neural network built to judge artwork adapted to drive an autonomous radio-controlled car
  • A neural network built to recognize a cat was trained to read CT scans
  • Neural networks built for other purposes can now find tumors in medical images much earlier than radiologists

Remember, the machines are learning how we learn, how toddlers learn. As professionals, our aptitude, knowledge and skills acquired through personal and business experiences is also transferable to other and varied domains, areas of study, business verticals, and enterprises.

This is the key to growth – learning, teaching, learning – as business leaders, as a society, and of course, as children, and now, machines do.

FRAMEWORK : Data Analytics

Simple Can Be Hard—for Machines and Humans

Who doesn’t want to simplify?

Search Google for advice on simplifying and you’ll find thousands of words of wisdom from Lao Tzu (6th century BCE) to Henry David Thoreau, to Mahatma Gandhi to Steve Jobs.

Today, machine learning is here to help us simplify, as Jongbin Jung, et al. explain in “Creating Simple Rules for Complex Decisions.”

In my own experience, the “art” is to know when simplicity is appropriate and when more complexity is required – and we must keep this in mind as we consider how machine learning can help us simplify making complex decisions.

The authors have “developed a simple three-step procedure for creating rubrics that improve yes-or-no decisions.” Their approach offers “the performance of state-of-the-art machine learning while stripping away needless complexity.”

The example they use – like the authors we discussed in FRAMEWORK (April 7, 2017) “A Guide to Solving Social Problems with Machine Learning” – is pretrial release decisions. By creating a rubric made of two weighted variables – age and past court dates missed – this simple calculation “significantly outperforms expert human decision makers.” The team created both weights for age and missed court dates and a “risk threshold” for the yes/no decision of release or return to custody.

Of course, oftentimes the path to simplicity is not simple.

What we don’t know is what data these variables are based on, how the leading indicators were created, what historical outcomes were studied, etc. Evidently, a lot of complex work went into creating this rubric.

The goal is always to create simple rules. However, whether it’s done by humans or machines, the process by which you create simple rules is complicated. For rules to be truly effective and simple, you must go through complex processes – and the machines will need our up-front help.

“Simple can be harder than complex: You have to work hard to get your thinking clean to make it simple. But it’s worth it in the end because once you get there, you can move mountains.”
—Steve Jobs