Machines for Navigating a Nuanced World

Jeff Woods


Whether or not you agree with the outcomes of the Brexit vote or the U.S. elections, in both cases, traditional models for predicting voter behavior clearly broke down. The factors that contributed to these electoral surprises indicate a profound shift in how people make decisions, with broad implications for enterprises.

Whether we aim to influence consumers to purchase a product or to guide employees to execute a new business strategy, we need a more nuanced understanding of human behavior and opinion. To give these their due requires us to consider more data and conduct deeper analyses.

To start, let’s consider what appears to have confounded some pollsters:

Digitalist Magazine Q1 17
  • Measurement errors: Analysts struggled to predict which voters would be most likely to go to the polls.
  • Undetected decision criteria: Voters surprised pollsters by choosing sides based on their wish to be in the winning faction rather than their perceptions of the costs and benefits of each position.
  • Culture jamming: Techniques for disrupting mainstream media with fake news, along with rising distrust in government and other institutions, challenged pollsters’ ability to pinpoint voter preferences.
  • Micro-targeting: The ability for campaigns to target each person with messages specially tuned to that person’s preferences and behavior made it difficult to ascertain the extent to which information broadcast to the general public influenced voters’ choices.

Each of these phenomena also affects how successful we are at modeling people’s behavior toward our businesses and in society at large. We need machine learning to help us derive a more individualized, behavioral view of the customers, employees, and partners that we are trying to influence.

We are just beginning to comprehend how to use machine learning, as well as to ascertain the practical and ethical limits to doing so. The quality of the algorithms we use to predict and respond to human behavior needs to improve. However, the initial advances indicate that machine learning, to a large degree, is the key to harnessing our understanding about how to influence people and their decisions in consumer, business, or social contexts.

That’s why we’ve dedicated this issue of Digitalist Magazine to questions at the forefront of machine learning in the enterprise. In “Empathy Machine,” our cover story, we explore the progress toward artificial intelligence that can read and respond to human emotions—the killer app for the digital economy. Our feature “An AI Shares My Office” cuts through some of the noise about work automation to uncover the ways that humans and machines will coexist in the workforce for the foreseeable future. We also look at the critical issue of how machine learning can help humans address the issue of bias—both conscious and unconscious—in their decision making. The feature “The End of Bias?” and our Thinkers interview with data scientist Cathy O’Neil probe the potential for substantially reducing the amount of unfair and unproductive bias in the world. We close with ideas from SAP chief innovation officer Juergen Mueller about how to apply machine learning across the enterprise.

Machine learning is ready to be treated seriously as a source of business opportunity as well as a management platform. We offer this issue of Digitalist Magazine to contribute to and advance that discussion.

About Jeff Woods

Jeff Woods is Chief Product Strategist, Products and Innovation at SAP.