7 Management Challenges for Mixed Human-Machine Teams [Ryan]

MANAGEMENT CHALLENGES FOR HUMAN-AI TEAMS

This chapter explores some of the challenges associated with managing teams of workers that include both humans (modified or unmodified) and machines of various levels of intelligence (smart, intelligent, or special purpose). Management is a complicated process that includes planning, resourcing, decision making, oversight, and process measurement. Executing management responsibilities necessarily means acquiring, organizing, training, and equipping the workers to do the job. In a world where humans and machines work together in teams, which means that these processes need to address the needs of both humans and machines — processes today that are done by separate groups of people with different training, skills, and job descriptions. This chapter is designed to help you understand how to begin to think about these challenges and how they fit into your life.

Note: the use of the term ‘artificial intelligence’ (AI) is problematic because of all the various definitions. In this chapter, no specific definition is used since it is well discussed in the other chapters.

STUDENT LEARNING OBJECTIVES

This is a philosophical exploration of a complicated topic. After reading this chapter, the student should explore these ideas further, particularly through the lens of life experiences.

PROLOGUE

A friend posited recently that good managers spend most of their time coaching, rather than engaging in the classical aspects of management (which include planning, resourcing, directing, measuring, and rewarding). In fact, she went so far as to opine that excellent managers spend about 70% of their time coaching. I asked her in response the following:

I wonder if you have considered what the effect of increased “AI”-human teams (where I use the term AI very, very loosely) might have on the role of the manager/coach? The emergence of a job that is (loosely) described as “AI prompting” might suggest that in these mixed teams, your formula will continue to be correct, albeit expressed in different ways according to the target. 

Her answer was, “With sentience, sure. What do you think?”

My reply:

Sentience is an interesting but unsatisfying measure. For one thing it is only indirectly measurable (via surrogate variables). But how do you detect if a thing is faking sentience? I think some mixture of sapience and agency makes a difference. The agency aspect is particularly interesting with regards to this challenge. 

Sentience, as a concept, is problematic as a measure of, well, anything really. It merely means the ability to perceive or feel things. Even plants can do that. How does that advance our understanding of future issues with AIs? Sapience is only slightly better, as it implies a measure of wisdom. Wisdom itself is a difficult concept to define, much less use.

But if you elide sapience into intelligence and add agency to that, then you get a being that has some level of intelligence that approaches wisdom and has the ability to act based on its own judgment. So this is the focus of the challenge: a machine that is intelligent in one or more aspect and which has the ability to decide if and how to act in any particular situation.  This implies more than simply following programmed logic: it implies the ability to collect data from the environment, use that data as input to a decision-making analysis, and then independently decide if and how to act based on the results of the decision-making analysis process.

So, here’s the thought: if you have two equally sapient individuals, each with agency, but one is human and the other is not, how do you manage them as a team? This is where we find ourselves. And it is worth considering how we manage today, how we will manage tomorrow, and how we should plan on managing in the future, assuming that humans remain in the management position.

The technology base of the modern economy is highly advanced in information processing and supporting functions. The people who work in this economy need to be highly skilled in order to use and work with intelligent machines. To date, these two components — technology and people — have been managed through separate systems. Until recently, that separation made good sense: people are different than machines and have different needs. However, there are emerging technology trends that suggest it is time to rethink this separation. As human performance is more tightly coupled with machines, including those classed as “artificially intelligent,” the approach to management needs to change. Human Resources (HR) and Machine Resources (MR) combine to create the modern workforce. Machines that collaborate with humans cannot be managed without taking the humans into account; concomitantly, humans that are tightly coupled with machines cannot be managed without taking the machines into account.

One of the interesting aspects of the thought exercise is considering how focused many managers are in removing or limiting the agency of human employees. Various human resources management techniques have been developed over time to reduce the ability of workers to act with a full, independent agency. These techniques include use of devices such as non-compete clauses, claw-back options for earned commissions, and psychological approaches to creating “firm as family” emotional ties. As a result, one of the potential futures that must be considered is a future where the machine has more agency than the human.

The purpose of this chapter is to explore that future specifically through the lens of management. You could just as easily change the lens to that of talent supply chain, innovation, or any other organizational issue. In fact, do so. It will enlarge your appreciation of the challenges ahead.

THE ROLES OF MACHINES AND HUMANS

Humans have always used technology. The first technologies were tools, clothing, and symbols. As new technologies were invented, merged with other technologies, and improved, the dependence that humans have on technology has grown significantly. We live in artificial caves that come with artificial sunlight and artificial heat. We cook on artificial campfires, clothe ourselves in artificial skins and cloth, and communicate using artificial noise. In truth, it would be difficult for a normal person to survive if plucked out of this cocoon of technological comfort and ease that we have created.

The most recent advancements of technology assistance to human endeavors have relieved humans of the types of work that humans do not perform well or consistently. This includes repetitious mindless activities, such as factory work, and large data set analyses. As humans have off-loaded these types of tasks onto machines, humans have been freed to focus on things that machines cannot do or do poorly. However, even as humans work in areas that machines are not (currently) good for, the execution of these types of activities relies in great part on technological support. And as tools like ChatGPT continue to make inroads into the workplace, the role of “technological support” becomes more closely coupled with human efforts.

In great nation competition, technology has been adapted and invented to protect humans from harm, both in offensive operations and in defense. Unmanned systems are paired with human controllers and weapon system operators. Intelligent targeting systems guide bombs to targets. AI assistants suggest and make decisions in ways that may be fully obscured from human understanding. The idea of “human in the loop” is being replaced by “human over the loop,” with the human taking on more of a role of coach or customer rather than participant.

The integration of technology into human existence has not been limited to performing tasks and serving as distancing mechanisms. It also includes integration into the human body for performance monitoring, maintenance, and adjustment. As noted in a  2012 study by the National Academy of Sciences, “Advances in medicine, biology, electronics, and computation have enabled an increasingly sophisticated ability to modify the human body.”  The study pointed to three general areas of technology innovation regarding the human body:  “human cognitive modification as a computational problem, human performance modification as a biological problem, and human performance modification as a function of the brain-computer interface.”[1]

THE NEED FOR COMBINED TALENT MANAGEMENT

In reality, combined talent management is already here, albeit in nascent form. Machines manage humans in all sorts of interesting ways, and machines are managed by humans. The various talents of human workers are judged and sorted by machines, while the various abilities of machines are integrated into workplaces according to need and potential to contribute.

During recent research into human talent management systems, a study group[2] discovered very interesting evidence of the integration of advanced technology into the process of managing humans. These integrated processes included the decisions associated with selection, assessing, and training the humans, sometimes to the extent of it not being clear how the humans would actually operate without the technology.  In other words, the machines were managing the humans, at least to some level. This suggests the potential of a future of human resource management that is more tightly coupled with machine management than we might be anticipating.

The adoption and integration of advanced technology, such as artificial intelligence, into the post-industrial economy, education, and governance structures have advanced the leveraging of knowledge to make processes more efficient and more effective. The field of knowledge management has attempted to capture the challenges of collecting, curating, and spreading both implicit and explicit knowledge within these environments, with some success. However, as the integration of knowledge, both as multipliers and as external replications of human intelligence, surges forward, there must be a concomitant recognition that organizations must deal with a supply chain of human intelligence — the brain power that humans bring to the working environment. The supply chain of human intelligence includes both the acquisition and value adding of intelligence in humans (hiring, training, and educating) as well as the integration of human intelligence into machines. Writ broadly, this describes a supply chain of intelligence that integrates the human and the machine. The supply chain approach can enable enterprises to value each contribution to the value-added prospect of the enterprise, enabling enterprises to account for investments correctly and adequately in the human intelligence supply chain.

It’s not just the cognitive processes of humans that are subject to supply chain like activities. The entirety of human and extra-human intelligence, either encapsulated in a human or as captured in a computer program, needs to be considered together. Humans use automation to outsource elements of intelligent activity. Alternatively phrased, humans extend their brain activity interaction with the rest of the world (other people, tools, places, etc.) by expanding their physical presence through the use of technology such as decision aids, AI, and other assistive technologies.

It is unlikely that the integration of human and technology components will cease, slow down, or reverse itself. In fact, we have seen direct evidence to the contrary, particularly as we have sheltered in place during the COVID-19 pandemic for an extended period of time, each of us learning new ways of collaborating with each other and socializing with each other. The future envisioned in The Caves of Steel[3] may be just around the corner.

Exploring the emerging issue of human-machine teams requires not just the investigation of human-machine teaming but also the supply chain of both artificial and organic intelligence. So, there are several management issues to consider. These include, but are not limited to:

• The behavioral and cognitive effects of human-machine teaming
• Issues of trust between humans and machine teammates
• Synchronization of the human and machine talent development and management systems
• The security of the “intelligence” supply chain

MANAGEMENT OF AI-HUMAN TEAMS 

When most people think of AI-Human teams, they reflexively think about data scientists or other high-tech people working with and developing AIs. A much more interesting team composition to consider is the ‘normal’ person teamed with one or more commercial production AIs.

Consider, however, examples of more ordinary work teams: that of a janitor, a lab technician, a kindergarten teacher, or a construction worker, each teamed with AI to accomplish their respective jobs. Each of those workers has specialty knowledge related to their jobs and presumably the AI will to. How will they ‘talk’ to each other? How will they coordinate? It is a manager’s job to make sure the working environment is appropriate to the needs of the worker (including disability issues).

A manager plans, resources, oversees, and measures the progress and success of work efforts. In order to successfully manage the human AI team, the manager (which may in fact be another AI) must take into accounts the differing needs of each member of the team. The manager hires (acquires) the talent needed, ensures each team member knows what they need to know to accomplish the job, provides the needed resources, makes the schedules, and assigns the activities. The manager also measures work performance, provides guidance on how each worker is performing, and integrates work output with other elements of the overall enterprise. It is challenging enough to do this well with only humans and dumb but critically important machines (like word processors, earth penetrating radar, or mass spectrometers). Adding what could be viewed as an alien being into that mix of personalities and technologies is very likely to result in novel challenges that are unpredictable right now.

BEHAVIORAL AND COGNITIVE ISSUES

One existing challenge associated with managing teams is observing, evaluating, and managing behavioral aspects of team relationships. We have names for different behaviors because these behaviors occur often enough to warrant consideration. Bullies, prima donnas, manipulators, shirkers, toadies, and more: each can cause challenges in team management. If left unmanaged, each can contribute towards the development of a toxic working environment.

What types of behaviors are likely to arise from the integration of sapient machines with agency? Will there be jealousies, attempts to sabotage the work of the AI, or other behavioral challenges with the human members of the team? Conversely, will the AI subvert the efforts of the humans?

Understanding that these need to be considered and watched for in the process of creating a true human-AI team is a critical first step to addressing potential problems.

ISSUES OF TRUST

Trust is a critical component of team efforts, particularly in work efforts that require a high degree of precision. When trust is lost between team members, the overall performance of the team suffers, sometimes to the point of becoming dysfunctional.

What does a human do if they lose trust in a machine they are working with right now? The answer is normally to try to fix it — go through trouble-shooting procedures, consult the tech manuals, ask the help desk, or reboot the machine. What will a human do in the future if the human loses trust in the AI teammate? It is entirely possible that AIs will be too complicated to trouble-shoot outside of an “AI hospital” and rebooting could do more harm than good.

What will an AI do if it loses trust in its human teammate? You can’t reboot a human, at least not today. Will the AI simply cease working with the human?  Will the AI create an environment where the human can’t successfully participate in the work processes? Will the AI damage the human in order to get it removed from the work environment? When trust is lost, it is normally for good reasons and the AI might feel like there is no option but to act in order to limit the ability of the untrusted (and untrustworthy) team member from working.

The issue of trust between team members then becomes a critical focus for managers of mixed AI-human teams. This issue becomes particularly acute when the loss of trust occurs during a fast-paced precision work situation. Consider an AI-human team operating on a human to remove a brain tumor. The loss of trust, by one member or the other, during that operation would be a potentially catastrophic problem.

THE TALENT SUPPLY CHAIN

Assuming a future where human intelligence is a desired attribute in the work environment, any Worker Resource (WR) system must consider the differences in the development of the various intelligences. For example, humans develop slowly; machines can be developed much more quickly. The needs of the work environment must be able to synchronize the time development schedules of both humans and machines in a way that optimizes the utility of each. The strategic planning issues associated with this synchronization are staggering, given the speed at which technology, and needed future human education can change.

Additionally, there are inherent security aspects of the human talent supply chain, even though it is not fool proof. The humans developed undergo years of training and education, during which they are observable and can be assessed for trustworthiness and competency. These same aspects are not inherently part of the machine intelligence supply chain, although they could be. The integration of parts manufactured in many different parts of the world, combined to create a body in which an AI can reside, creates a universe of supply chain issues associated with the trustworthiness of parts — both as individual components and as elements of systems.

FINAL THOUGHTS

As the previous chapters in this book have demonstrated, we are headed into a future that will provide many challenges. Thinking about how humans fit into that future is an important task. Research, obviously, is needed. Another thing that is needed is a profound sense of caution. Embracing the promise of technology is critically important but it must be done with clear-eyed testing, analysis, and adaptation of the technology to ordinary human existence. With such an approach, the stars become ours to conquer.

ENDNOTES