Confronting AI’s Ethical Dilemmas

In this new and changing world where data collection is never-ending, the laws surrounding the collection and handling of the big data it generates are still being written.

shutterstock 503725969 1280x1280
Dell EMC

Oh, the wonderous things artificial intelligence, machine learning, deep learning, big data analytics, and the correlations and causations from data science are uncovering these days! With so many new findings, it is critical for CIOs to stay on top of the ethical challenges these new insights are bringing their way.

Data monetization is a growing big data analytics business, anticipated to reach over $200 billion by 2020. The result of such analytics is expected to expand the growth of data from 33 trillion gigabytes (33 zettabytes) in 2018 to 175 zettabytes in 2025. All this data being analyzed and sold means more and more businesses are going to know more about each of us. Let’s look at some examples…

Pattern recognition

As unique as each of us would like to believe we are, we fall into patterns of other like-minded, like-health, like-hobby, like-style, like-habit people … hundreds of thousands of them. When we check in, share a new product we purchased, or share an opinion or a photo, we are openly giving data to others to collect. Based on the insights found from such analytics, what obligation does the chief information officer have to customers?

It was just about eight years ago that Target broke the wonderful news of the impending delivery of a new grandbaby to an unaware parent of a teenager whose shopping patterns matched those of others who were pregnant. This story raised a topic of discussion in nearly every university about what ethical obligations businesses have when the data about their customers’ behaviors reveals something that perhaps the customers didn’t want others to know.

No doubt, the creepiness level of a company sending prenatal advertisements to teenagers is pretty high. However, that was in 2012. With today’s much greater volumes of data and insights, where has the standard shifted? For example, if a person’s shopping, activity and posts collectively indicate a very high probability they have a life-threatening illness, is the organization identifying the probability under an obligation to share this insight?

Facial recognition

Executives and other business-level decision makers may have their eyes on amazing new revenue streams that facial recognition can generate. However, a lack of accuracy and lack of standard of handling of such data means those responsible for the collection, storage, analysis and destruction of this data are balancing on a slippery slope of ethical decisions.

For example, using facial recognition data for security purposes originally sounded like a great idea to government agencies and others working to catch criminals and keep people safe ― that is, until it was found that identification of ethnic minorities had an up-to-35% error rate. Consequently, cities on both coasts have decided not to use it. However, many others do use it.

It’s important to bear in mind that, when facial recognition data is collected, ethical data storage decisions must be made, such as

  • Is long-term storage acceptable for images of children without their parent’s consent?
  • If security footage of an event was recorded, how long should it be kept?
  • If a retailer identifies a customer with video and verifies it as that person via their credit card transaction, and if the video identifies logos on their clothing, is it okay to send targeted advertisements from those logos’ organizations to them?

Ethical frameworks

Fortunately, CIOs need not create their own ethical frameworks while trying to wrap their heads around these latest technologies. John McClurg, Vice President and Ambassador-At-Large for Blackberry Cylance, has identified five AI ethics frameworks to get technology and ethics leaders started.

  1. Asilomar AI Principals
  2. IAPP - Building Ethics into Privacy Frameworks for Big Data and AI
  3. The IEEE Global Initiative on Autonomous Systems
  4. Universal Guidelines for Artificial Intelligence
  5. EU Councils Guidelines on Artificial Intelligence and Data Protection

What may be the most important concept specified in these frameworks is that they differentiate “ethical” from “lawful” behavior. In this new and changing world where data collection is never-ending ― as devices are always hearing, always seeing and always analyzing ― the laws surrounding the collection and handling of the big data it generates are still being written. However, placing the value of people first will never be an out-of-date concept.

To learn more

For more perspectives on unlocking the value of data with artificial intelligence systems, explore Dell Technologies AI Solutions and Dell EMC Ready Solutions for AI.

Copyright © 2019 IDG Communications, Inc.