Cold but Fair? Algorithmic Management in the Workplace
Algorithmic management promises efficiency, scalability, and data-driven decisions. It often delivers on that promise. The problem is not what algorithms do, but what they typically don’t do. They don’t listen. They don’t explain. And they don’t necessarily answer.
Text by Martti Asikainen, 5.5.2026 | Photo by Adobe Stock Photos
Algorithmic management has been spreading rapidly. In an OECD employer survey (2025), 90 per cent of US companies reported using algorithmic management systems. In Europe, the figure stands at 79 per cent. In Finland, nearly one in two specialist and knowledge workers reports having direct experience of machine-driven management (Asikainen & Lahtinen 2025).
There are good reasons for machine-driven, or algorithmic, management systems. They scale, they don’t tire, and they don’t have favourites. When an algorithm measures performance, it measures all employees against the same parameters. This is a promise that appeals to many. It is no surprise, then, that algorithmic management crops up daily in shift scheduling, performance measurement, and task allocation.
An algorithm has no interest in which ice hockey team you support, how many employees it has to shepherd, or what time of day it does so. Its output is not dependent on mood, nor is its operation affected by health problems or internal organisational schisms. Efficiency and fairness are two different things. A system can be one or the other, but rarely both at once. Then again, much depends on how it has been built, for what purpose, and in whose interests.
Three Stumbling Blocks
The research literature on algorithmic management is extensive but fragmented. No unified theory has yet emerged, and different lines of enquiry frequently talk past each other (e.g. Kellogg et al. 2020; Meijerink & Bondarouk 2023). This fragmentation is not merely a theoretical problem — it is also reflected in practice, where algorithmic systems are deployed without any settled understanding of their effects.
Three recurring themes do emerge from the literature, however. The first is a lack of transparency. Algorithmic management works at its best like a good line manager: consistently, predictably, and with reasons given. At its worst, it operates as a black box in which decisions are produced but their logic remains hidden. The academic literature stresses that openness alone is not enough — what is decisive is comprehensibility and explainability (Ananny & Crawford 2018; Lee et al. 2015). An employee’s ability to grasp the system’s logic has a direct bearing on their engagement and trust (Rosenblat & Stark 2016).
The second theme is psychological strain. In research, algorithmic feedback appears both as a motivating challenge and as a paralysing obstacle. The critical factor is the individual’s sense of control — are you directing your own work, or are the algorithms directing you? Research on short-term and gig work repeatedly documents stress and a persistent sense of being evaluated (Wood et al. 2019), but also situations in which clear feedback and goals increase motivation (Kellogg et al. 2020). Perceived fairness emerges in the literature as a key mediating factor. A system regarded as fair generates less strain regardless of its technical characteristics (Colquitt et al. 2001; Meijerink & Bondarouk 2023).
The third theme is behavioural change. University of Oslo professor Taina Bucher describes the phenomenon as pacifying the algorithm — a situation in which people learn how the system evaluates them and begin optimising their behaviour according to its logic (Bucher 2018). In practice, this manifests in adjustments to working patterns, such as underpricing oneself and self-censoring one’s behaviour (Carnegie 2022). The same pattern has been observed in both platform work and more traditional organisational settings (Kellogg et al. 2020).
![[Infographic: A quadrant showing the respective strengths of algorithmic and human management. The horizontal axis runs from 'difficult to measure' to 'easy to measure' decisions; the vertical axis from low to high need for human interpretation. Top-left ('Human strong'): conflict resolution, appraisal conversations, staff wellbeing. Top-right ('Hybrid management'): performance reviews, recruitment, bonus models. Bottom-right ('Algorithm strong'): shift scheduling, resource allocation, cost optimisation. Bottom-left ('Risk zone'): situations where the system operates as a black box with no interaction or means of challenging decisions.]](https://www.fairedih.fi/files/2026/05/Gemini_Generated_Image_kxex7hkxex7hkxex-1.png)
The Obvious Solution Is Not Enough
Given these three problems, it is tempting to reach for the simplest explanation: open up the algorithms, make them transparent, and the problem is solved. The EU AI Act (2024/1689) moves in this direction, requiring high-risk systems — including those used in recruitment and human resources — to be explainable and auditable. Regulation is necessary, but it is unlikely on its own to foster understanding, build trust, or change behaviour.
Ananny and Crawford (2018) drew an important distinction on this point in their article Seeing without knowing. Seeing, they argue, is not the same as understanding. You can be given access to an algorithm’s code, its weights and parameters, and still be entirely baffled by why it made a particular decision about you. Transparency without interpretation and context is rather like receiving your medical records from a doctor written entirely in Latin.
Amazon provides a telling example. Its systems are documented, audited, and in many respects explained. Even so, employees describe the environment as oppressive and closely monitored — some have compared the surveillance culture in the workplace to that of a prison (Palmer 2020; Sainato 2024; Bansal 2026). Wood et al. (2019) have shown that algorithmic management brings with it both autonomy and control, simultaneously and within the same system — and this double-edged quality does not disappear as transparency increases.
The question, then, is not simply one of openness, but also of power. Who has the right to challenge a decision? Who has access to the data? And who is accountable when the algorithm makes a mistake? In an OECD survey published in 2025, nearly two in three managers using algorithmic management reported at least one concern about the reliability of their systems. The most common concern related to an unclear locus of responsibility in situations where an algorithmic system had led to an erroneous decision.
When responsibility is distributed into a system, it does not disappear — it becomes invisible. A manager can point to the system as having made a decision, but the system itself can neither defend itself nor say a word in response. The employee is easily left entirely alone with a decision whose origins cannot be traced or attributed to anyone. Opening the source code will not fix this problem; it requires genuine human engagement.
When an employee receives a decision they do not understand — a cancelled shift, a drop in performance scores, a contract termination — they are not primarily looking for an explanation. They are looking for a person to whom they can say: this is not right. An algorithm cannot hear that. It cannot hesitate, apologise, or change its mind. It is present everywhere at the level of data, and absent precisely where experience is formed. This is problematic because management is, at its core, about encounters — situations in which two people negotiate what is fair and what is not. When one party is replaced by an algorithmic system, the negotiation ceases, and what remains is simply a decision one must live with. This is why finding the right balance between algorithmic and human management is especially important.
![[Infographic: Three key challenges of algorithmic management. First: opacity — decisions are visible but their logic remains unclear. Second: psychological strain — the persistent sense of being evaluated increases stress and reduces the feeling of control. Third: behavioural change — people begin adjusting their behaviour to fit the logic of the algorithm. The image's message is that understanding these challenges is a prerequisite for fair and humane algorithmic management.]](https://www.fairedih.fi/files/2026/05/Gemini_Generated_Image_fwb4o8fwb4o8fwb4.png)
Cold but Fair
In a survey conducted last year by Haaga-Helia University of Applied Sciences, Finnish workers gave algorithmic management surprisingly positive assessments (Asikainen & Lahtinen 2025). More than 1,700 specialists, white-collar employees, and managers responded, of whom nearly half felt that a machine could be a fairer manager than a human being. 38 per cent believed that algorithmic management would have a positive effect on career development (Valkama 2025). The most important justification, however, was not technical but social.
Many respondents saw the machine as basing its decisions on data rather than on personal relationships. This finding says something essential about how poorly human management has succeeded in many organisations. When employees express a preference for a machine over their line manager, this is not a case of technology fetishism but of disillusionment with a working life that harbours inner circles and old boys’ networks, opaque criteria for promotion, unfair feedback cultures, and chronic understaffing. It is little wonder that many younger workers in particular experience systems that place everyone on an equal footing as more just than a human manager.
This finding is consistent with what Rosenblat and Stark (2016) documented in their study of Uber drivers. They showed that the algorithmic system created an illusion of apparent fairness — a kind of impartiality that attracted workers, even though the information asymmetry in reality always favoured the platform owner (e.g. Partington 2026). Trust in an algorithm, then, does not always mean the system is fair; it may simply mean that human management has betrayed the worker’s trust in one way or another.
Many of those who responded to the Haaga-Helia survey nonetheless clearly prefer a human manager to an algorithm. Respondents were unwilling, for instance, to have a machine make decisions about setting and evaluating performance targets. When it came to equitable pay, however, the machine was seen as at least the equal of a human manager, or even slightly superior. People trust the machine’s impartiality, then, but fear its coldness and its capacity to evaluate human conduct.
At root, this is a question of conscious division of labour. Algorithms excel where data is dense and criteria are readily measurable. Human beings, by contrast, shine when situations call for interpretation, context, and empathy. The challenge lies in the fact that organisations do not always draw this distinction — or alternatively, it is made under financial pressure, almost without anyone noticing, and without anyone ever really asking what we want the algorithm to do, why, and what it should under no circumstances be permitted to do. Then again, perhaps the fundamental question is not whether the algorithm can be trusted, but whether the organisation deploying it can.
![[Infographic: A continuum between human and algorithmic management. Left: 'Warm but unfair' — inner circles, opaque promotion criteria, unfair appraisals. Centre: 'Balanced zone' — a conscious division of labour between human and algorithm. Right: 'Cold but fair' — risks of algorithmic management: continuous monitoring, scoring, decisions with no possibility of negotiation. The image's central message is that the goal is not to choose one extreme but to build a system that is both fair and humane.]](https://www.fairedih.fi/files/2026/05/Gemini_Generated_Image_lxzsovlxzsovlxzs.png)
About Author
Martti Asikainen
Communications Lead
+358 44 920 7374
martti.asikainen@haaga-helia.fi
This article was produced as part of the RoboBoss and FairAI projects, funded by Haaga-Helia University of Applied Sciences and the Finnish Work Environment Fund.
References
Ananny, M., & Crawford, K. (2018). Seeing without knowing: Limitations of the transparency ideal and its application to algorithmic accountability. New Media & Society, 20(3), 973–989.
Asikainen, M. & Lahtinen, A. (2025). Algorithmic management spreads across Finnish workplaces — younger workers show greater acceptance than their older colleagues. Press release. Haaga-Helia University of Applied Sciences. Published 16 June 2025. Accessed 26 April 2026.
Bansal, V. (2026). Amazon is determined to use AI for everything – even when it slows down work. Published in The Guardian, 11 March 2026. Accessed 24 April 2026.
Bucher, T. (2018). If… then: Algorithmic power and politics. Oxford University Press. Oxford.
Carnegie, M. (2022). Gig Workers Are Getting Crushed by the Review Mill. Published in Wired, 15 December 2022. Accessed 24 April 2026.
Colquitt, J. A., Conlon, D. E., Wesson, M. J., Porter, C. O., & Ng, K. Y. (2001). Justice at the millennium: A meta-analytic review of organizational justice research. Journal of Applied Psychology, 86(3), 425–445.
European Union (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council on artificial intelligence (AI Act). Official Journal of the European Union.
Kellogg, K. C., Valentine, M. A., & Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366–410.
Lee, M. K., Kusbit, D., Metsky, E., & Dabbish, L. (2015). Working with machines: The impact of algorithmic and data-driven management on human workers. Proceedings of the CHI Conference on Human Factors in Computing Systems, 1603–1612.
Meijerink, J., & Bondarouk, T. (2023). The duality of algorithmic management: Toward a research agenda. Human Resource Management Review.
OECD (2025). Algorithmic management in the workplace: New evidence from an OECD employer survey. OECD Publishing. Paris.
Palmer, A. (2020). How Amazon keeps a close eye on employee activism to head off unions. Published on CNBC, 24 October 2020. Accessed 28 April 2026.
Partington, R. (2026). Dynamic pay on platforms such as Uber should be banned, says TUC. Published in The Guardian, 3 May 2026. Accessed 3 May 2026.
Rosenblat, A., & Stark, L. (2016). Algorithmic labor and information asymmetries: A case study of Uber drivers. International Journal of Communication, 10, 3758–3784.
Sainato, M. (2024). ‘You feel like you’re in prison’: Workers claim Amazon’s surveillance violates labour law. Published in The Guardian, 21 May 2024. Accessed 23 April 2026.
Valkama, H. (2025). ”Ei ennakkoasenteita eikä kaveripolitiikkaa” – Tekoäly nähdään ihmistä reilumpana johtajana. Published on Yle, 23 September 2025. Accessed 26 April 2026.
Wood, A. J., Graham, M., Lehdonvirta, V., & Hjorth, I. (2019). Good gig, bad gig: Autonomy and algorithmic control in the global gig economy. Work, Employment and Society, 33(1), 56–75.
