Yves here. Please welcome Thomas Klikauer and Meg Young, who delineate some of the implementations of computer and spyware-enabled Taylorism that they call algorithmic management. It’s been distressing to see the tacit acceptance of management using technology to monitor workers and enforce relentless and generally unreasonable performance demands. Given how tight labor markets have become, one can hope to see some employers marketing a traditional (as in free of tech-enabled harassment) workplace to attract staffers.
By Thomas Klikauer, who has 700 publications and writes regularly for BraveNewEurope (Western Europe), the Barricades (Eastern Europe), Buzzflash (USA), Counterpunch (USA), Countercurrents (India), Tikkun (USA), and ZNet (USA). His next book is on Media Capitalism (Palgrave) and Meg Young (GCA and GCPA, University of New England at Armidale), a Sydney Financial Accountant and HR Manager who likes good literature and proof reading
Increasingly,algorithmic management is used to control workers using software technology– the technology of mathematical algorithms. Management’s algorithms that control workers can best be understood as opinions expressed in a mathematical formula.
In the realm of management, these become managerial operations formulated as equations put into a computer-coded program for management. These formulas are based “on” management and they operate “for” management. They do exactly what management tells them to do. Algorithms are not neutral technologies. Instead, they are tools that serve management.
Algorithmic management that controls workers is mostly used in the vicinity of line management. Primarily, algorithmic management is a technique to remotely manage, discipline, and punish workers. Essentially, it relies on two things: data collection (e.g. big data) and the remote but ultra-tight surveillance of workers. Put into a formula, an algorithm allows management to run automated or semi-automated decision-making processes and surveillance operations.
Unsurprisingly, big data collection and mathematical algorithms are central to algorithmic management as they allow for the functioning of digital labour and other online platforms. In turn, digital labour platforms are ideal for the harvesting of data on workers to be used against them.
Online work and platform work are linked to the so-called gig economy which has roughly two parts: in the first part, workers offer online services on a platform like cleaning, data entry, or even selling products; in the second, and more common part, workers are employed by an online platform like Uber.
The second form is entering ever more areas of work such as, for example, one of the most commonly used, is that of warehouses (Amazon). But algorithmic management is also found in retail, food delivery, manufacturing, marketing, consultancy, banking, hotels, call centres, and even among journalists, lawyers and the private police, security where online platforms like life eye surveillance offer management ever more control.
In these jobs, algorithmic management means the automatic allocating of work tasks to workers through handheld or wearable devicelike smartphones and computers. For example, the so-called ride-hail-driver are forced to turn on their smartphone app to receive the next work task. On it, workers receive algorithmically organised trip and delivery requests from management.
Workers are given a 15 second time window to accept or reject the next so-called “gig” – a work task. If they accept the request, which they mostly have to do in order not to be punished and disciplined or lose their jobs – workers are then given a delivery or passenger’s location via the management app’s map display.
Algorithmic management automatically withholds key work information such as for example, fare and destination. Yet, management platforms also restrict the ability of workers to decline trips allocated to them algorithmically. Not surprisingly, 80% of all workers accept management’s tasks allocated to them automatically. They tend to do so mostly likely out of fear to be disciplined. As one worker said, you do what the App tells you.
All this also means that algorithmic management does not fully replace real management. Instead, there is a continued relevance of real managers and line supervision. This comes not “despite” but “because of” the managerial use of algorithmic management. Essentially, algorithmic management delivers new tools to control workers but the controlling is still overseen by management. For one, managers are needed to oversee and supervise dispatchers and to monitor the monitors that control individual riders and delivery drivers, as in the case of Uber, for example.
In warehouses, algorithmic management runs control largely through the so-called handheld devices, also known as scan guns.These link barcode scanners, motion detectors and location tracking devices while establishing near total control over workers. And as if this is not enough, the Coronavirus pandemic has given management an even more ammunition. In some cases, management has installed cameras in warehouses that operate machine learning systems. These notify management when a worker violates social distancing rules introduced, overseen, and enforced by management.
For example, a Driveri is an AI equipped camera used in warehouses and delivery vehicles of some so-called outsourced delivery service partners. It is a form of faked employment based on the hallucination of workers as independent contractors. In any case, such spy-on-workers cameras capture not just the road ahead, the workers, and both sides of the vehicle, it also instruct and control workers.
It can force workers to take almost any managerially requested actions in response to what management deems to be a violation of its rules. This can mean not to engage in unplanned stops such as for example, a toilet break, a sandwich on the way, etc. it may also mean to take a managerially requested break if the AI spy camera has captured a worker yawning.
A so-called optimising software call Percolatallows algorithmic management to ensure the most profitable mix of workers for maximising sales in every 15 minute slot of the day. Meanwhile, Genome enables algorithmic management to calculate the average time it takes for a worker to complete a work task. It can also alert team leaders and bosses when work is deemed to be behind schedule. Similarly, with software like RescueTime, algorithmic management can reduce work distractions that management deems to impact negatively on workers’ productivity and, hence on corporate profits.
Worse, behind all this lurk platforms that use customer ratings, rating and ranking acceptance rates and evaluate workers algorithmically. This turbo-charges service capitalism’s prime ideology of the customer is king to almost semi-dictatorial levels – every customer becomes a kind of a boss.
In hospitality, online ratings are often issued on the whim of customers making highly subjective reviews. Yet, management gains those from powerful websites like TripAdvisor.Worse, they are incorporated into workers’ individual performance reviews. They also inform managerial team meetings in which decisions on workers are made. As a consequence, workers understand quickly that they are under constant surveillance from customers and their impulses. Meanwhile, call centres programmes such as Cogitocontrol workers in pretty much the same way. A worker noted,
Talking too fast? The program flashes an icon of a speedometer, indicating that he should slow down. Sound sleepy? The software displays an “energy cue,” with a picture of a coffee cup. Not empathetic enough? A heart icon pops up.
At Amazon, data collected from workers’ hand-held devices are analysed to automatically rank workers from highest to lowest under what management knows as rank & yank.Donald Trump calls it, you are fired!It is also known as, my way or the highway. If workers are lucky, the lowest 10% are told to speed. This can still be done by real managers.
The evilness of the managerial system was perfected by Jack Welch’s maxim, fire the bottom 10%. Shortly before that, a line supervisor may walk by showing a worker a print-out making clear that he is below 90% while telling the worker, you have to work a bit harder. The next step is the my-way-highwayprinciple.
Of course, algorithms are a great tool for management not only when dishing out disciplining actions against workers. Almost self-evidently, line managers and supervisors make great use of an automated metric as well as an automated ranking algorithm. They even prefer them when applying their own discretion because unbiased-looking algorithms powerfully support what they do.
Algorithms aid the appearance of neutrality – it is all in an algorithm, a technique, a machine, a mathematical formula. This makes it easy to justify management decisions and it makes it easy to convince workers, these algorithms – not management – made the decision. The perfect smokescreen for managers.
It remains imperative to understand that algorithmic management reshapes organisational control. Algorithmic management does not eliminate control. Instead, algorithmic management enhances control. While algorithmic management takes out some managers, it does not eliminate management altogether. Instead, it gives management new tools to control worker ever more tightly and rigidly.
Of course, algorithmic management is used by management to intensify work, to increase monitoring and to enhance surveillance, to raise work paces, to minimise even the tiniest gaps in workflows, and to extend work activity well beyond the conventional workplace and working day.
In some cases, management requires workers to work at frantic pace. At Amazon, workers often have to run in order to keep up with the speed. In a well-known case, a worker was forced to pee into a bottle, as Amazon boss Jeff Bezoadmitted.
Overall, the level of algorithmic management differs from company to company with bottle-peeing Amazon on the lower end. Yet, the Society of Automotive Engineers (2014) introduced a useful categorisation of automation that can be applied to algorithmic management (AM). It has six levels:
- No AM: standard management prevails – no algorithms;
- Assistant AM: algorithmic management is used to support management;
- Partial AM: parts of management are done by to algorithms, real managers are decision-makers;
- Conditional AM: most management decision are handed over to algorithms;
- High AM: most of performance management, control, and worker discipline is handed over to algorithms;
- Full AM: algorithms define performance, evaluation and control without the involvement of real managers.
Whatever the form algorithmic management takes, it is likely to enhance management control over workers. Ever since Richard Edwards, we know that there are three forms of management control:
- Simple control is based on direct control by a visible and present authority figure like a supervisor or line manager strictly enforcing the top-down supervisor-subordinate hierarchy.
- Technical control relies on some physical devices like machines, an assembly line, computer software, algorithms, etc. It moves from face-to-face control onto technical control by machines.
- Bureaucratic control uses rules such as HR-policies to control and manipulate workers often laid out in handbooks, training programmes, policies, performance management, etc.
In this framework, algorithmic management may well fall into the second group of technical control because algorithmic management relies on a technical apparatus – an algorithm – to control workers even though real managers are not totally removed as the list of the automotive engineers shows.
Set against this is the fact that algorithmic platforms can not only be used to control workers. Online platforms can also be used by workers to eliminate management control. Unions and union-based organizations such as, for example, the CWU, the Tech Work Coalition, Coworker, Unionbase, the widely known Change.Org, ADCU, etc. call this platform organizing.
Platform organizing uses three forms of union organizing: a) labour and solidarity organizing by a highly-skilled and core workforces mostly of tech companies; b) resistance by workers at gig companies that apply algorithmic management to control their labour forces using digital platforms; and the c) unions develop counter-platforms for workers dedicated to facilitating communication among workers for the purposes of labour organizing.Overall trade unions recommend ten demands when management introduces algorithmic management:
- workers need to have a right to be informed about algorithmic management and its potentially harmful effects on workers;
- the rule that a human manager always overrides an algorithm remains the overarching principle;
- algorithms have to support human managers and not replace them;
- transparency by management about algorithms remains essential;
- unions need to be informed about the program management is buying and its algorithms need to be assessed on potential health risks (e.g. work stress, etc.), bias, and discrimination;
- management needs to clarify and justify which data on workers are collection and which surveillance regimes are planned;
- workers and union representatives are to be given access to all data collected by management about them and about the forms of algorithmic performance assessments that are conducted;
- benefits from algorithmic management (e.g. productivity, greater flexibility, etc.) need to be shared with the workforce on equitable terms;
- whenever management plans to introduce algorithmic tools, it should have a people planoutlining parallel investment into the workforce (e.g. training); and finally and perhaps most importantly,
- before introducing algorithmic management, employers should justify why it is actually needed at all.
In the end, algorithmic management is not going away any time soon. Instead, the very opposite is the case, algorithmic management is spreading. With this come rafts of pathologies that turn workplace increasing into dystopian places to be avoided. Yet, trade unionscan play an active role in shaping even those workplacein which algorithmic managementhas been introduced.
For algo management, you don’t need people, you need robots. Of course, people are still cheaper than equally agile robots (for example), so Amazons and similar of the world will try to get the cheap people..
I’m going to venture that for management, algorithmic management can be great. So a lot of controversial decisions can now be done by these algorithms instead of managers with no effect on one’s career. If there is any problems that arise because of them, you just say that it was the algorithms and that it was out of your control. If there are ensuing legal problems, who are they going to take to court. The algorithm? For a lot of managers, it is like being able to eat great, big deserts with sprinkles and a cherry on top but that the algorithms get all the calories.
I’ll leave it to the IT mavens here to comment whether its connected to this topic, but I recently had a conversation with a colleague about a very expensive IT system in her sector which is being (quietly) shut down. The reason is apparently (so I can gather from a few sources), the system was intended to mix internal record management with external accessibility to information, but flaws in its original design means that it is too easy for hackers to penetrate.
The core problem seems to have been that a system designed from the beginning with the priority of tracking and controlling internal work systems was incompatible with wider organsational aims and legal requirements. I have heard from two relatives of mine in IT consultancy that all too commonly they are requested to provide systems with a focus on work monitoring, but management won’t listen to arguments that this can create multiple problems further down the line when the systems are refined for outside access or internal useability. Or put another way, what seems smart for managers at the time can turn into timebombs for organisations.
That’s very interesting. From my experience as a programmer, I can tell you that it is very difficult for management to put into English what the alorithm is actually doing.
When I first saw the title of this article, I thought it was about managing algorithms, not about algorithms managing people.
It’s very difficult to make sure that the computer is doing what you think it is doing.
Most managers are not techies, and most managers are loathe to try to explain things to techies, so how do they verify that the programs really are generating the results the sales people are promising to generate?
Managers will not be able to just wave their arms and say, “The computer did it”. They will have to give the raw input to the plaintiff’s counsel, and allow the plaintiff’s counsel to generate the same output.
I bet most will not be able to do that.
Interesting days ahead.
You bring up an interesting point- managers in an AM regime don’t require much in terms of skills. The algo tells them what to do, if it doesn’t actually do it automatically. Low skill management means lower pay packages.
I’d like to see that, but, I think that will take a decade to get there. Courts, law enforcement, and the entire legal system, are IT- illiterates to an astonishing and frightening degree. We have: Law enforcement who are very excited by data, yet easily fooled by the same data, lawyers (all sides) and judges cannot comprehend how and why this is possible even with expert support.
That decade will be spent effectively by lobbyists like The Singularity Institute on entrenching “Digital Management” in the entire public sector, we hear too much about “digitalisation ready legislation”, and of course in the management of legislative process, thus tilting the odds in favour of “the algorithm” winning against the people.
We already see how it may go down in the way that NPM makes the IRS and the police favour simple crimes made by small people. Their driving KPI’s, I assume, being something like “Cases vs. Successful Prosecution” and “Low officer-hours per case”.
A complex case, like proving systematic discrimination performed by an AI, will end up a long way down in the priority queue. Unless it is really obvious, but, the whole idea with using the AI to f.ex. undercut discrimination laws is to make it non-obvious and out-source-able.
September 22, 2021 at 7:13 am
and Margaret Bartley
This was in Links this morning.
It seems that no one can put into English what the algorithm is actually doing.
In Cory Dotorow’s “Eastern Standard Tribe” some of the protagonists works as high-powered executive business consultants, ala McKinsey, except, they are in reality an organisation of anti-corporatists and anarchists: They deliberately get corporations ensnared in perfectly crafted smart tech solutions, that will get their tentacles into every nook and cranny of the corporation, and then blow them up a bit down the road!
There is all the talk of labor shortages – the real shortage is in leadership and management – not quantity (way to infested) but quality – there are enough incompetent and poor leaders and managers to sink all the raised boats in the world -… they already have.
Of course the less capable and more incompetent the ownership and leadership is….the more dependence upon the app band aide to hide this fact from everybody else. The old emperor has no clothes deal.
I am lucky to have worked for some really great leaders, owners and managers and know a bunch more but the other set, a majority group of them, are much less able or even give two sh#$S about their management style or competence and just roll along consuming their time with ways to prevent the more competent and able underlings from taking their jobs – turf protecting themselves behind an app wall.
I guess I am a cynic
And then we have the argument for returning to offices that the informal meetings (breaks) between workes/colleagues are essential for companies.
The level of monitoring written about here is not new to the people who worked in call-centers 20 years ago, it has just spread to more environments. What was the saying again about: First they came for… So yes, what was once done to call center workers are now being done to warehouse-workers and shock horror also to the office-worker.
I doubt that it is cost-effective, it is done because it can be done. Cost-effectiveness is probably only possible if costs can be externalised, either to the worker or the wider society. Break the worker and have someone else pay for the healing of the worker.
It might be possible to survive such monitoring long-term if the work-day was only four hours long. Eight hour days in those conditions is likely to break many, if not most, if done for years.But shortening the work-day can’t be done because reasons.
I can easily imagine how algorithmic management will play an increasing role in the operation of large medical clinics/chains …
It will likely look something like this.
I recently watched Idiocracy again and there was a throwaway line in there about being fired by algorithm – clearly it was a movie ahead of its time. And it didn’t take 500 years for us to get someone like President Camacho either, only about ten.
I tried to find the clip about being fired by algorithm on youtube, but after a few tries it didn’t come up and I had to settle for the one linked above. Clearly youtube’s algorithms could use some work. Maybe they can hire some AI to do it crappily for them…
I take issue with the ‘customer is king’ assumption, at least how widespread it may be. That is not what I have observed and heard about in board rooms and industry conferences.
The appropriate mantra is ‘owners are king’ for many companies, as execution of the business model is paramount. Do you really think that those PE models and consulting decks are ignoring that? They know that dividends and paydays are the route to owner success. Some feel-good customer talk is marketing but doesn’t show up as readily as articles wish.
That the customer is used in such ways does not bode well for long-term company health, and is just one of many factors that may explain many current problems. At least the customers get noticed, unlike the employees.
There’s also these sort of more cuddly in appearance but still pernicious HR algos for generic office workers. I work in “AI” / natural language processing and can say with conviction that the stuff does not “understand” language to the extent that any really nuanced automatic snooping is possible — but… blanket surveillance isn’t done because it ‘works’, but more for the vibe it creates.
The usual PR move for these kinds of interventions is to center a nice, possibly socially beneficial use case. Here it’s little reminders when composing an email that you might be wording something in an insensitive way (in a different arrangement of power at work, this isn’t wholly terrible). In other cases it’s stuff like “scanning ur tweets for signs of depression, phoning a shrink immediately” — which I believe is something Apple is making a push for recently, but the PR strategy has been floating around in the academic side of the field for years.
Going unsaid are more likely use cases like… detecting labor organizing / unrest.
Yes, indeed, it seems like communist regimes look pretty weak compared to this new regime (uh, I don’t have a name for it).
With wonderful ideas like algorithmic management it does seem strange that businesses everywhere are having trouble finding workers. Maybe those Luddites had the right approach after all.
Simon Head has published two great books on this topic: ‘The New Ruthless Economy’, back in 2003, and ‘Mindless’ (2014). Highly recommended.
As if the managers are not slaves to the algorithms as well… as if the algorithms haven’t replaced many of them already… as if the corporate behemoths that employ such technology are even mildly democratic, humanistic, or compassionate… this is the logical outcome of neoliberalism (capitalism?): ‘individual’ (distinct from human) as machine.
I often explain, that the problems the world faces are deeply philosophical, but I fear that by the time anyone realizes, there will be no humans to speak of… perhaps some derivative thereof… algorithms?
The coup that was Citizens United did not elevate corporations to the status of personhood, but lowered persons to the status of corporation. A resounding denial of the soul…
Senior execs at JP Morgan certainly didn’t like being watched by AI. https://www.bloomberg.com/features/2018-palantir-peter-thiel/
“It all ended when the bank’s senior executives learned that they, too, were being watched, and what began as a promising marriage of masters of big data and global finance descended into a spying scandal. The misadventure, which has never been reported, also marked an ominous turn for Palantir, one of the most richly valued startups in Silicon Valley. An intelligence platform designed for the global War on Terror was weaponized against ordinary Americans at home.”
Peter Thiel has to be in the top ten worst people in the world.
“benefits from algorithmic management (e.g. productivity, greater flexibility, etc.) need to be shared with the workforce on equitable terms”
I don’t know why, but when reading this bit I can hear someone laughing hysterically somewhere.
What happens behind a warehouse wall is nobodies business but the warehouse wall owner, am I right? No speed limit to the grinding abuse under Bezos’ whip hand.
Amazon shopper = whip cracking sadist