Nearly half of the workforce in the United States raised concerns about the potential threat of artificial intelligence (AI) displacing their jobs, as revealed in a survey conducted in February by the investment banking firm Jefferies.
Beyond the realm of banking, managers are also urged to consider the implications of AI encroaching on their roles.
According to researchers at ESMT Berlin, AI presents an opportunity to oversee research projects, enabling operations at a scale and efficiency surpassing human capabilities.
Maximilian Koehler, a PhD candidate at ESMT, and Henry Sauermann, a professor of strategy at ESMT, advocate for the integration of AI as a supervisory tool in a paper titled “Algorithmic management in scientific research.”
Published in the academic journal Research Policy (Volume 53, Issue 4, May 2024), the paper is accessible via SSRN without any subscription barriers.
The authors posit that AI-driven tools can enhance human productivity by expediting tasks such as reviewing scientific literature, formulating research inquiries, aiding in data analysis, and predicting novel drug compounds. However, they emphasize that AI cannot entirely replace human expertise—at least not presently.
“While AI’s capabilities as a ‘worker’ have advanced significantly, human scientists will remain indispensable in the foreseeable future, as the scale and complexity of research endeavors continue to expand,” the authors assert. “Therefore, we shift the focus from AI as a worker to exploring its role as a ‘manager’ overseeing human researchers engaged in scientific tasks.”
The authors highlight numerous instances where algorithmic management could optimize productivity across various research projects.
“The evolution of artificial intelligence has empowered AI to substantially elevate the efficiency and scope of scientific research by orchestrating intricate, large-scale initiatives,” stated Koehler in a press release.
To evaluate the feasibility of algorithmic management, the researchers scrutinized around 200 research projects to assess how they addressed five key managerial challenges: task allocation, direction, coordination, motivation, and fostering learning.
Following extensive interviews and investigations, they pinpointed 16 projects and two platforms that incorporated automated management to varying degrees.
These initiatives include Aurorasaurus, Crea.visions, eBird, EteRNA, and Galaxy Zoo, among others.
For instance, in Galaxy Zoo—a collaborative project for classifying galaxies—the AI system is tailored to enhance participant engagement, resembling gamification strategies used in platforms like Uber to retain gig workers.
“The AI predicts the likelihood of participant disengagement and intervenes with messages to boost user motivation,” the paper elucidates. “It strikes a balance between timely interventions that could disrupt workflow and delayed messages.”
Although this form of algorithmic management did not prolong user engagement with Galaxy Zoo, it notably accelerated the classification process without compromising data quality.
Drawing comparisons with projects devoid of managerial AI, the authors contend that those leveraging such mechanisms tend to be larger and associated with platforms, benefiting from shared technological infrastructure. This, they suggest, carries implications for platform dominance and the strategic approach of major research institutions like universities towards research funding and IT infrastructure.
The researchers advocate for further exploration into the ramifications of algorithmic management, underscoring the distinctions between scientific work and gig or office-based employment, which have already been subject to studies on the impact of algorithmic interventions.
“While autonomy has traditionally been a fundamental aspect of scientific endeavors, valued greatly by researchers, algorithmic management may encroach on this autonomy if AI monitors individuals continuously,” the authors remark.
They caution that these systems introduce ethical and legal dilemmas concerning the exploitation of motivational mechanisms and the control of worker data related to their skills, motivation, and performance. For instance, the sharing of worker metrics derived from AI management systems could influence future hiring decisions involving these individuals.
Nevertheless, even if AI assumes some managerial responsibilities, it does not signify the demise of traditional managerial practices such as lavish lunches, golf outings on expense accounts, inflated contracts, favoritism, and arbitrary directives to subordinates.
“If AI can streamline routine management functions, human leaders could redirect their focus towards strategic and interpersonal tasks like identifying high-impact research objectives, securing funding, or cultivating a conducive organizational environment,” Sauermann remarked.
Alternatively, they could also wield their authority to restructure organizational dynamics through staff reductions.
