The Future of Work Podcast

Hello and welcome back to another ILO Future of Work podcast.

I'm Sophy Fisher.

Innovations in technology have always affected the world of work,

but the issue is back on the news agenda and seems to have taken

on a new urgency with the rise of AI or artificial intelligence,

in particular with the launch of ChatGPT last autumn.

Commentators are divided on the effects this generation of AI could have.

Some are predicting it will cost millions of jobs.

Others say it could actually support decent work by helping

workers and taking over some of the more mundane tasks.

Now, I'm being joined today by Virginia Doellgast,

Professor of Comparative Employment Relations at Cornell

University's School of Industrial and Labor Relations.

She has been at the ILO to take part

in the Regulating Decent Work conference,

which has covered the effects of AI on decent work.

Virginia, thank you for coming.

Welcome to the studio.

Thank you for having me.

Let me start first of all by saying, AI is not actually new, is it?

I mean, we've had it around for decades

and it's actually already embedded in a lot of our daily lives.

We just don't notice it.

What's so exciting now?

That's right, I've thought about that as well

because I've been studying AI for a couple of years now

and it has been around for a while in the workplaces

that I've gone into, a lot of these tools

are pretty well established and of course,

as you said, in our daily lives, AI is all around us.

It's used in our translating tools or web searches as well as of course

behind the apps, we use to hail taxis or-. -Analyzing images.

analyzing images.

-[crosstalk] Imagery. -Exactly.

There have been a lot of books written by academics in the popular press

looking at the job destroying effects of automation,

coming from AI or how AI is used in recruitment

or algorithmic management kinds of tools.

I think

there has already been a public discussion,

but I think that the actual way that these tools could be used

and their impacts were relatively abstract before.

The recent impact, as you say, is connected to this wide accessibility

of ChatGPT and other tools for generating images

where the public has this kind of tangible,

tactile experience with these tools.

I say I've been presenting my research

on AI for the past two years

and when I was pre ChatGPT,

you often had a response, "Oh, are these tools really that good?

Will they really have any impact on jobs?"

Also, just a lack of understanding of what the technology was.

I've seen a huge change in even recent months when I present

some of this research, everybody is much more excited

and has more of an intuitive sense.

Now you find yourself working on one of the hottest topics

in world of work academic research.

As I understand it, there are different approaches to looking

at how digital tech and AI is changing our world of work.

There's economic, there's social, and you have a third one

which is called CER, comparative employment relations.

Tell us a bit about your approach.

Comparative employment relations, basically,

I think what's distinct about that or the way that I study work

and other colleagues in my field is that I think

maybe to back up and think about, how do economists study these tools?

Economists tend to estimate the impacts of AI-based

automation on jobs and skills and tasks.

You have a lot of sociologists

and legal scholars who have done more case study based

work or looked at legal implications of algorithmic management tools

for fairness, for monitoring intensity, these kinds of things.

Employment relations really has a distinct focus

on how these different kinds

of AI-based tools are used together at the workplace level,

within firms, within industries,

and also trying to understand then how institutions,

both collective bargaining institutions through unions,

but also national institutions and legal frameworks

together are influencing how these tools are adopted,

as well as being able to influence and change

the impacts of those tools on work and on workers.

Great. Of course, from the ILO's perspective,

that's exactly what we are interested in.

Now, your research has focused mostly on the ICT industry,

communications, information technology.

Why did you choose to look at that industry?

An important reason to look at the ICT industry.

When we think of this sector, it encompasses telecommunications, IT,

big Tech and game development, a lot of different segments,

but across this industry,

this is really where AI-based tools are being developed.

A lot of firms in these sectors are really early adopters.

They're on the forefront of the developments and use of AI,

but also you have very diverse segments and jobs within this one sector.

You can study the impacts on higher-skilled

jobs like programming and engineering,

as well as more easily rationalized or automated jobs like call centers,

back office, you think about game testing work,

which is where you have a lot of union organizing right now.

Across these jobs in these industry segments,

union presence and worker representatives,

you have them in the traditional telecoms,

they have a long history of unionization,

but then you have a lot of new organizing that unions are doing in the Tech sector.

Within one industry, in other words,

you are getting pretty much a look at everything.

You're getting super high-skilled cutting edge.

You are getting much more basic repetitive tasks,

and you are also getting the larger traditional forms

of employment with contract and everybody in one place,

and then what is often referred to now as gig,

where people are are very dispersed and working by themselves,

and negotiating their own contracts.

You've got everything in one industry?

That's true. You have everything in one industry,

and you also have a sector that has a lot of experience

with actually developing these tools.

Some of the workers in the sector

are actually the programmers who are making the AI.

[laughter]

There's an irony in there somewhere.

Let's go back to your approach, your CER approach.

From that perspective,

how do you see AI changing the way work is organized?

To some extent, obviously, it follows on from that, ultimately regulated.

The impacts of AI on work,

I think of them as falling across a number of different areas.

I've described them in the past as labor replacing,

labor controlling, or labor displacing, but another way of thinking

about that is just the impacts on job numbers,

skills, intensity, discretion, location.

Can I just get you to unpack those three for a second?

Labor controlling would be, give us an example of that.

Labor controlling is more around algorithmic management.

In other words, an algorithm checking

your work just to make sure you are doing it,

and how often are your productivity rate.

What about labor replacing?

That would simply be somebody losing their job to an algorithm, would it?

Yes, or it can also be replacing certain tasks or certain skills.

Which of course might be bad tasks

that you don't want to do [?] I'm saying this.

Labor displacing, what's that?

Labor displacing is more about the location of work.

Remote working.

In other words, you don't have to employ somebody in California anymore.

You could employ somebody in Kolkata?

For an example, call center workers can now be working from home.

Increasingly, with the pandemic, the COVID-19 pandemic,

you saw companies moving a lot of workers to work from home contracts,

and those really were enabled

by the ability to use more AI-based monitoring, in their home workplaces.

Thank you. Sorry for that little digression,

but I thought we ought to explain it for the listeners.

Let's go back to the point you were making

about how you see AI changing the way that work is organized.

Can I give an example from call centers?

Please.

Because I think this makes it more concrete.

If you think about that first set of effects,

as labor replacing, you see chatbots

which everybody's now had experience with.

These are basically peeling off simple calls,

and these are increasingly enabled by AI using

some similar technology to chatGPT,

and then you have robotic process automation,

which is basically automating a lot of follow-up after calls

that agents used to be doing online before they would take the next call.

You have these front-end assistance

where agents can then find customer information,

or it will lead them through.

It's a little prompt on their computer screen that's acting

like a chatbot and telling them what they should be doing

to help them answer difficult questions.

These tools make workers more efficient.

It makes the work more intense.

You imagine you used to be doing a lot of different things,

now you're just handling calls

all the time and you're handling more complicated

calls because all the simple ones are being handled by the chatbots.

All of that will change the kind of content of work,

make work more intense and have these mixed effects on skills.

On the one hand, workers are getting more complex calls,

but then one worker can handle a wider range of call types.

On the other hand, they're not having to use

their skills as much because the tools are leading

them through all the other kinds of skills.

It has a lot of implications.

Then on the algorithmic management side, the worker controlling,

increasingly the call center worker has a coaching app that's AI-enabled,

that's going through and telling them,

"Oh, you need to speak with more empathy to your customer," for example.

Of course one person's coaching app is another person

looking over your shoulder all the time, isn't it?

Yes. It's another form of intensified monitoring.

It could be a tool, it's helping me to do my job better,

but it also then can be used by management

to have a lot more information about what the customer's doing.

Presuming, this issue will be that if this job is becoming

more stressful and hopefully more productive,

is that additional productivity and stress being reflected

in the workers' remuneration benefits?

That's something we're going to have to tackle

as well, isn't it?

Again, it's a bit of this question of how the companies are evaluating

what the impacts are in skills, for example.

Also in cases where they have unions,

are they renegotiating things like job classifications

and is the union willing to do that?

That will have some effect on how these jobs are organized?

Certainly, I've already seen

in these telecom firms in different countries

that employers are starting to come to unions and saying,

well, some of the skills are changing.

Maybe we need to change the compensation structure because of that.

Or the unions are coming back and saying, this work is more intense.

Workers are handling more complicated calls,

so their bonuses need to reflect that.

They can't handle them as quickly.

Yes, because if you're only getting the complicated calls,

the calls can't be dealt with in 30 seconds.

They're probably five minutes

which might actually look bad for your productivity bonus.

I don't just want to focus

on what might be perceived as the negatives.

There must be positives for workers

in additional AI in their working regime as well, yes?

Yes. I've already talked about skills which could also be a positive aspect.

It's important aspect to emphasize that AI can often be seen

as a tool not only for high-skilled workers

but also for well, call center workers can also be very high skilled.

For example, we did a survey of call center workers

in the US and we asked a number of questions

about different kinds of AI-based software

that they were using in their jobs.

We found that they were actually quite positive where they saw

these different software tools or applications

helping them to do their jobs better, to improve customer service

because they care about doing a good job.

They like it when they can find the information easily.

Okay. That's one way that it can improve positives

to workers and presumably also it means that high-level jobs are available

in more geographically diverse locations.

You no longer have, for example, to move to Palo Alto in California

or to Mumbai to get a high-skilled job.

You can remain somewhere where perhaps you have family connections

or you have some reason that you can't move

and still get a higher quality, hopefully higher paying job.

That's true. It provides more job opportunities.

Also to rural areas, it means that workers don't have to commute

or to change locations as they're searching

for different kinds of jobs.

It improves mobility from the worker side.

You could also say there are ways

in which these tools help reduce management bias,

could potentially help reduce bias.

If you think about it, for example,

AI is increasingly being used in tools like scheduling

or training to recommend training or to help organize scheduling.

That takes away some of the control for managers who might have their own

human biases and it could create a more transparent tool.

For example, in Germany, I talked to some works counselors

at Deutsche Telecom who were developing a scheduling tool that gave

individual workers much more control and flexibility, which was using AI.

This was something that was then designed together with the works council

and management that made sure

that it built in a lot of controls and safeguards.

Okay. Well, that's interesting and it brings me

onto the next question I wanted to ask you, which is,

what are the implications for workers' organizations

and in particular trade unions?

Significant. Yes.

The examples that I just gave, I think about the positive

and the potential positive and negative aspects

are a way of thinking about how there is this huge range

of possibilities of how you're using these tools and really figuring out

how to use them in ways that both are positive for employees,

but also are positive for productivity or for customer service.

These other kinds of outcomes we want.

You really do need worker input and worker voice

particularly with new kinds of technologies.

They're often buggy, they make mistakes.

The workers on the front lines are the ones

who see those and can help correct them,

or can help figure out how to make them work in more efficient ways.

You need unions on the one hand to put

negotiated limits to make sure that managers are not using these tools

in a way that significantly degrades

the quality of work, intensifies monitoring,

is de-skilling work, but also to negotiate,

to make sure that they're used more as tools

to compliment skills to empower workers,

and firms can really benefit from that as well.

Presumably, now, some of these AI tools are so complex

that actually it's pretty hard for a manager,

particularly one who doesn't have a PhD in a relevant

discipline to actually understand what they're doing,

how they're doing it, and what the consequences are.

They actually need more assistance to convert

something that's theoretically a tool of help

into something that actually improves things

on their shop floor or whatever it might be.

Yes, that's right.

I talked to a consultant who was working

with call center managers in many different countries,

and these were non-union workplaces, and he was saying,

that he's brought in because people buy these speech analytics

kinds of AI-based tools

and then they just try to put them in place

without thinking about their whole broader management policies

or how they're actually going to use this data and information.

In a sense, they're helping management to build in something

like employee voice into the design of these tools.

A real advantage of unions and works councils

is that you have that resource

to figure out how to design them in more efficient ways.

I can see a challenge here simply with the issue of organization,

which is fundamental to workers moving together,

be it in a union format or in any other,

because if AI as you said is encouraging work displacing,

people are working at home.

They may not even be in the same time zone.

They may not even speak the same languages.

Also, some of these people will be young, people with high-tech skills,

high ICT skills, who aren't used to the concept of working

within an organized labor structure.

There's obviously a big challenge here for unions

actually to get workers together so they can speak with one voice.

That's true. It is a big challenge.

You see a lot of organizing often starts

in these different areas of the tech industry

where either you've had a tradition of unions like in the telecom sector,

but then a lot of the traditional telecom unions

in many different countries are now expanding out

and starting to have different organizing projects in these newer segments.

They are having success.

Often, for example, in game development,

it starts off often with the workers who have maybe

the most easily rationalized

or the jobs with more intense conditions like game testing.

Increasingly, you see more solidarity developing broader work together

between union organizing within companies like Alphabet

which is the Google parent company.

Then you see them doing a lot of advocacy and publicity around trying

to improve conditions down the value chain for people who are doing,

for example, AI coding and AI testing.

There's a lot of talk now about regulation of AI,

how do you see the challenges of that?

Do you think it's actually going to be possible to put

any kind of shape or form on how this industry develops?

I think that yes,

it is possible to shape how the industry develops

but as you say, it is changing quickly.

On the one hand, I think it's a good idea to have broader regulations

that set a floor of certain hard limits on how AI is used

to make sure that it's not used in ways that could potentially harm

both employees but also the general public.

Some basic tools like the GDPR,

the General Data Protection Regulation in Europe

is certainly an important tool that covers a broad range of uses.

The AI Act also sets, I think, reasonable ground rules.

There are efforts now in the US also to think

about what some of those ground rules would be

in ways that don't stifle innovation,

but that also protect workers in the public.

Then, on the other hand,

I think just having better regulations to support unions

and to support collective worker voice are really key.

That means codetermination rights or bargaining rights,

but also basic freedom of association.

This is the ILO.

That's something that in my own country,

the US is not always so well enforced.

This means making sure that unions can be present,

can organize and more firms and have the rights to have information

about and to negotiate over these tools.

As you just mentioned, in addition to the traditional areas of concern,

like freedom of association,

and voice and the right to collective bargaining,

this also leeches into areas like privacy,

like creativity, and intellectual property.

It's a very, very complicated area.

It is, that's true.

A lot of those regulations in a sense.

I was talking to one expert in the game development industry.

He was saying that he felt that

what was really going to drive decisions about AI use,

and a lot of decisions on programming

or the way in which generative

AI was being used, was going to be through lawsuits,

because of intellectual property.

You already see different countries trying

to figure out how they're going to regulate

intellectual property for that reason.

You also see unions negotiating over this.

I don't know if you follow the Writers Guild strike in the US.

You had negotiations where the union was trying to put

into their contracts, basically,

a prohibition against training AI on content

that was generated by the writers themselves.

In a sense, this is a copyright issue.

Yes. There's also been a case with somebody,

a voice artist who voiced up one item and has now found that his voice

through AI is being used for all sorts of subsequent products,

with which, of course, he is not getting paid and never gave permission.

At the time that he wrote the contract,

it could never have been anticipated that this would have been a use.

You think that regulation

and the shaping of the industry will have to come

from a whole different range of things ground up

and top down from multilateral institutions,

and also from these areas of use utilities, such as legal cases?

Yes, I think, definitely, it will.

I think that unions, but also, basic other civil society organizations,

social movement organizations will all be involved, and making sure

that there's both a worker and a public voice in these decisions.

Again, to emphasize the importance of having

a clear framework for labor rights,

so that you have this flexible ability of workers within workplaces,

firms, and industries to respond.

Coming back to where employment relations connects to all of this,

my focus has clearly been on that worker voice piece.

I've seen very much in my comparative research between Germany and the US,

that where you have in Germany a framework,

where you have rights to negotiate, for example,

they have clear codetermination rights on how technology

is used for behavior and performance control,

which is a very clear right that we don't have in the US.

It means that there's a lot more possibilities

to actually come up with jointly negotiated standards that meet both labor

and management's needs in that area.

Once again, the importance of workers' voices comes to the fore.

Let's leave it on that note.

Very important note for the ILO,

very important note for the development of AI.

Virginia, thank you so much for your time.

That was Virginia Doellgast who is a professor

of Comparative Employment Relations

at Cornell University in the United States.

For now, let me wish you all goodbye.

Please join us again soon for another edition

of the ILOs Future of Work podcast.

Goodbye.

[music]