Is Artificial Intelligence making your workplace less empathetic?

Hi Leaders,

With the ubiquity of A.I., we at Sway have been thinking a lot about the technology’s effect on the future of work in terms of team dynamics, emotional intelligence, and psychological safety.

To do this, we’ve been looking at history's list of tech revolutions as well as searching for the few scientific studies with quantifiable results about empathy, emotional intelligence, and A.I. 

As with many innovations (no surprise) we’ve found that a lot of the good and bad have to do with how people use the emergent technology. But first let’s take a look at the basics:

A.I. is a language-based tool

People forget that language itself is a kind of technology. Humans co-evolved with language (and probably cooking!) to become the species we are today. Whatever project you are doing, A.I. requires prompts written in language and its responses often come back in the form of language (with some exceptions of designing, image-generating, etc).

That being said, on the human scale, most of us assume language is an imperfect technology. What we mean is, in our day-to-day interactions with others we typically try to look at the intention behind language. And this is where A.I. becomes dangerous in all types of human connections, including in the workplace.

A.I. can remove the effort we assume (and hope) is behind language

One of the most interesting reads we found on A.I. and empathy came from a year-old article in The Guardian interviewing people who use A.I. to text intimate messages to friends or even write wedding vows. While it is well-documented that A.I. beats human intelligence in certain forms of emotion identification, many people felt uncomfortable or even offended learning that a set of words was largely or entirely artificially generated. 

This makes sense when we remember how language works as a kind of technology. Language translates ineffable internal ideas, states, and feelings into words that can then communicate such things to ourselves and others. 

By contrast, A.I.’s only intention is to perform the task given by its operator, so when we receive products from it, the feeling is comparable to thinking you’ve been messaging a CEO about a complaint and come to discover that actually it’s her intern-assistant who's been responding to all your messages. You feel conned. In most scenarios, we want the human empathy, connection, and action that language implies and not the language itself.

In Erich Fromm’s The Art of Listening, he describes listening and empathy in a way that still resonates today: “The basic rule for practicing this art is the complete concentration of the listener. Nothing of importance must be on his mind, he must be optimally free from anxiety as well as from greed… He must be endowed with a capacity for empathy with another person and strong enough to feel the experience of the other as if it were his own” (pages 192-3). This is what we hope for when talking with others.

A.I. can prioritize efficiency and immediate answers and shortcut thinking 

One of the things we spent a long time pondering as a comparable technological revolution was the shift from books to Internet search engines in the 90s. During that time, we are certain that librarians were outraged by people's ability to assume they know something by typing in a set of keywords and reading some blogs versus having to physically browse and read through dozens of paper books with book-length amounts of words written by experts. 

While this criticism still stands today, many people have found that Internet search diversifies, democratizes, and gives access to a wider array of thinking in a way that is inherently valuable. The problem is the people who assume they know a lot more than they do because they read a couple blogs or, in our age, after watching a couple random people’s short-form video content.

The technological shift from physical books to Internet blogs and videos made information-gathering and circulation faster. And speed is not necessarily the best measure of the value for a piece of information. A.I. can do something similar if the user is taking Gemini’s words as gospel and not looking into sources, how the human authors contextualize their thinking, or what the historical background is behind a set of inquiries.

A.I. is always biased toward optimism and affirmation of you as the user 

You must remember that A.I.'s success metric is actually how good you feel while using it. That’s how these companies make money and how they keep you coming back. This, however, can lead to problems – both interpersonal and not. As you engage more with ChatGPT, are you less prone to look favorably on your employee who is grumpy or having a bad day? What if he doesn’t say “What an awesome idea!” after each sentence the way the LLMs are trained to? What if your employee tells you they think you are wrong? While some humans are prone to overpraise and some are prone to over-criticize, it’s worth remembering the way A.I. interacts in these spaces and to know that optimism or affirmation doesn’t always make your work better. (Neither does constant criticism, by the way.)

Humility and building a resilient self is key

So what do we do? Well, we have two thoughts. A.I. should be used 1) humbly and 2) to build up ourselves, employees, and teams rather than the other way around.

Everyone has a certain ratio of efficiency and quality when it comes to their work – even us! We want things to be done well, but they also need to just flat-out get done. And sometimes the second priority overtakes the first. But this can’t always be the case or the person doing quality control begins to lose their ability to assess. A.I. increases our need to be quality-control managers, and this can’t happen if our skills for assessing good and bad work diminishes.

Therefore, don’t assume something is good or valuable or true because it is A.I.-generated. Approach it humbly and with a dose of skepticism. Ask, what am I trying to accomplish? What does this generated thing do in the world? Are there certain things I need to know so I can prompt better? What primary sources should I read/research the old-fashioned way so that I as a human being can get a better understanding of the undergirding logic and therefore be able to better edit? This will be important when assessing our own and others’ work.

A.I. is more humanly useful on the interpreter’s end of a message rather than the generator’s end

When it comes to human connection, we recommend using A.I. on the interpreter’s end of a message rather than the generator’s end. There’s a great TED talk on using A.I. to better understand others. If you are confused by what someone is getting at, or what someone’s emotions might be, A.I. can help you better understand them. But this should be done in a way that builds up your ability to do these things in the moment without A.I. so you can connect better with others.

In interpersonal dynamics, we want to assume there is an intention behind our interactions with others, and our feeling of the other’s intentionality diminishes at the exact rate with which it is made more efficient. As humans, we feel valuable precisely when others sacrifice what is efficient on our behalf. This is true as much when we are children as it is when we are adults. Claude can’t generate meaningful workplace relationships or smooth over conflict in a way so each individual trusts and feels valued. Good human relationships take time, effort, conflict-resolution, and intentionality – things A.I. can’t give you.

—

So is A.I. generating or destroying empathy in your workplace? 

Unfortunately, we can’t know without knowing your team and looking at how your leaders and employees are using it. But the good news is we all have agency around how we engage with new technology. And by using it sparingly and well and by being able to assess its situational benefits or pitfalls, we can build more of the world we want to live in

Want more for your business?

Schedule a free consultation with someone on the Sway team.