Published on: June 7, 2021
Authors: Richard A. Warters
Ripped straight from the headlines; “...Drivers Launch Legal Action Over 'Robo-Firing' by Algorithm”, “Staff 'hired and fired by algorithm'”, and, “ (Airline) is using an algorithm to select who should be fired”. The termination of employment is the working world’s equivalent to capital punishment. When exercised en masse, corporations are accused of exercising corporate greed. When exercised without warning, especially by machine, the conditions are ripe for industrial martyrdom.
It comes as no surprise that unions are taking a particular interest in the application of Artificial Intelligence (AI) as the world struggles to govern 21st century gig economy workers under mid-20th century labor legislation. Last year, in anticipation of the European Commission’s white paper “On Artificial Intelligence - A European approach to excellence and trust”, the ETUC insisted: “AI - Humans must in be command” as they called for:
“…an ambitious European AI strategy to maintain and reinforce workers’ protection and involvement, and empower them to create sustainable use of AI tools.”
As I began thinking about this, I defaulted to, “Humans are in command.” In a 1988 Information Technology course, I used the BASIC programming language to build an accessible job application system. It could have been installed in a kiosk and people could apply electronically.
That is a given today. Not so in the days before the World Wide Web. The tech was rudimentary, but at its core were the rules of the developer - me. My simple tool dot matrix-printed pages of “IF-THEN” statements to make “decisions” based on the rules I set. If today’s date minus your date of birth did not total to 18 years of age, the process stopped. Not a U.S. citizen? No need to continue. (Those were simpler times.)
Today’s AI is a long way from my 1988 employment tool, but many of the fundamentals are the same. Computers still operate on the basis of zeroes and ones. If-then statements establish the foundation of many programs. And the developer’s rules still establish the initial outcomes. But AI goes well beyond. In “What is artificial intelligence?”, the Brookings Institute outlines three definitive criteria for AI; intentionality, intelligence, and adaptability.
These are the conditions that enable our car’s navigation system to redirect us around delays we can’t see. At the core, human beings determine the rules, establish the logic, and do the math. But the outcomes – the turn-by-turn directions - are based on a set of constantly changing conditions. Real-time traffic congestion, construction delays, and even our own driving habits are considered and processed by the millions of lines of code that chart our course.
We will get to the destination we set for our daily commute, but the route we take on any given day may differ as often as there are roads that lead in the right direction. Pop culture has already worn thin the joke about the driver who turns because his “NAV” said, “Turn right, now.”, only to end up in a pond. We should not blindly accept the directions of our navigation systems. Nor should we expect our employees or their representatives to blindly accept the output from an employer-initiated algorithm that could result in the loss of one’s livelihood.
The Brookings’ authors offer, “As long as there is little contentiousness or disagreement regarding basic criteria, these systems work intelligently and effectively.” Unfortunately, the basic criteria – the rules - are the crux of the contention between employers and their workers and/or their representatives.
Moreover, the rules define the bounds of the relationship and establish expectations for workers. I believe Employee Relations (ER) is uniquely charged with identifying ways to connect people with their work in increasingly productive ways. The most fundamental piece of the ER puzzle is to ensure set performance expectations are clear at the outset of the work relationship. That is the first and most critical connection.
We can drop workers into a pasture and let them wander blindly into the electric fence that is discipline. Keep them guessing where the boundaries are and pretty soon, you’ll have a workforce of people (1). too timid to act, and (2). ready to leave. They certainly won’t be increasingly productive.
Alternatively, we can ensure expectations are clear, how work will be monitored, and the consequences, good and bad, of performance and non-compliance. If AI is employed as part of our work-monitoring protocol, sound ER practice demands workers and/or their representatives understand the performance expectations built into the program, the inputs it will receive, and intent of the logic – whether legally obligated or not.
AI, and the algorithms that make it possible, are the latest, and admittedly, maybe the most powerful tools for the supervision of work yet. But they are just the latest in a line of shiny new objects to cloud the relationships between supervisors and the supervised.
We can choose to get hung up on an algorithm that few of us in HR and as few of “them” in unions would understand. And we can argue the proprietary nature of the code developed by the wunderkind in his or her parent’s basement. Or we can agree to share the intended outcomes of the data we are capturing and processing in the name of efficiency and the maintenance of an inclusive and productive work environment.
Exercising sound ER practices will require that humans are in command in the employment decision-making process and in the exercise of due process. The road to hell is paved with good intentions.
By definition, AI, in conjunction with machine learning and data analytics, enables intelligent decision making and adaptability. We need to constantly ensure employment-related outputs reflect the original intent and that decisions are being administered in a way that ensures fairness and equity for our workers, whether they are “employees” or not.
Letting a machine – however shiny and new – determine discipline unchecked would take a load off of supervisors and ER practitioners’ plates alike. But we are a long way from the ability to quantify every variable of the work experience into ones and zeroes. Fairness and equity are at the core of my number one operating principle of Industrial Relations; Respect the People.
You can unleash AI, but you’ll need to do the math and show your work – the expectations, inputs, and intentions – if you want to connect people with their work in increasingly productive ways.