Stanford Analyzes Worker Preferences For AI

Many of us have internalized this notion that we’re soon going to be working side-by-side with robots, or at least AI agents and entities.

So as humans, what do we want these digital colleagues of ours to do?

How does delegation work?

A Stanford study recently went into this where authors surveyed 15,000 workers in over 100 types of jobs, to see what they really thought about AI adoption.

Why They Did It

I thought this comment by one of the authors sums up the purpose of the report well:

“As AI systems become increasingly capable, decisions about how to deploy them in the workplace are often driven by what is technically feasible,” writes project leader Yijia Shao, a Ph.D. student in the Stanford computer science department, “yet workers are the ones most affected by these changes and the ones the economy ultimately relies on.”

In other words, it’s the front-line workers who are going to be most affected by these changes, so we might as well hear what they have to say (in addition to doing all kinds of market research.) There’s a reason why the suggestion box is a time-tested element of business intelligence. Technology has to be a good fit – it’s not something you just implement carelessly, throwing darts at a wall, and then expecting all of the people involved to sign on and go along for the ride.

Some Results

In terms of actual study findings, the Stanford people found that a lot of it, as Billy Joel famously sung, comes down to trust: 45% of respondents had doubts about reliability, and a reported 23% were worried about job loss.

As for the types of tasks that workers favored automating, the study provides a helpful visual that shows off various must-haves against certain danger zones of adoption.

Specifically, Stanford researchers split this into a “green light zone” and a “red light zone”, as well as a “low priority zone”, and an “opportunities zone” featuring uses that workers might want, but that are not yet technically viable.

Uses in the green light zone include scheduling tasks for tax preparers, quality control reporting, and the interpretation of engineering reports.

Red light uses that workers are wary of include the preparation of meeting agendas for municipal clerks, as well as the task of contacting potential vendors in logistics analysis.

There’s also the task of researching hardware or software products, where surveyed computer network support specialists seem to prefer to do this type of work themselves.

I thought it was funny that one item in the low priority zone was “tracing lost, delayed or misdirected baggage,” a job typically done by ticket agents. It explains a lot for those legions of hapless travelers entering their faraway AirBnBs without so much as a toothbrush.

As for opportunities, it seems that technical writers would like AI to arrange for distribution of material, computer scientists will largely sign off on technology working on operational budgets, and video game designers would like production schedules automated.

Why Automate?

I also came across a section of the study where researchers looked at reasons for automation desire on the part of survey respondents.

It seems that over 2500 survey workers want to automate a task because it will free up time for other kinds of work.

About 1500 cited “repetitive or tedious” tasks that can be automated, and about the same number suggested that automating a particular task would improve the quality of work done.

A lower number suggested automating stressful or mentally draining tasks, or those that are complicated or difficult.

The study also broke down tasks and processes into three control areas, including “AI agent drives task completion”, “human drives task completion” or “equal partnership” (and two other gradations). You can see the entire thing here, or listen to one of my favorite podcasts on the subject here.

One of the headline items is a prediction of diminishing needs for analysis or information processing skills. That connects with more of a focus on managerial, interpersonal or coordination job roles. However, how this will shake out is concerning to many workers, and I would suggest that 23% of respondents worrying about job displacement is a wildly low number. Almost anybody anywhere should be worried about job displacement. Regardless of what happens in the long term, many experts are predicting extremely high unemployment in the years to come, as we work out the kinks in the biggest technological transformation of our time.

Anyway, this study brings a lot of useful information to the question – what do we want AI to do for us in enterprise?