Thursday, June 2, 2016

Leadership in the Age of Robots

In March of this year, a Google computer defeated the world’s reigning Go champion in four out of five matches. Several years earlier, IBM’s Watson computer defeated two of Jeopardy’s greatest champions and IBM’s Deep Blue competed successfully against he former world chess champion Garry Kasparov. We are all aware of the use of robots in hospitals and combat zones, and the evolution of autonomous cars.

The military uses robots for several purposes, including the recovery of improvised explosive devices. Hospitals are also increasingly using robots for guiding patients and delivering drugs. More recently, as reported in the Economist (2016), researchers have created “robodocs,” robot surgeons that successfully stitched up the intestines of piglets with a minimum of human supervision. Under a surgeon’s supervision, the Smart Tissues Autonomous Robot (STAR) was able to sew piglets’ guts together after the doctors had severed the piglets’ intestines. In fact, STAR was able to carry out about 60% of the procedure without human intervention and its stitches were more evenly spaced and the sutured guts less leaky than what surgeons would have done.

There is no question that we are in the midst of a new age of artificial intelligence and the use of robots. Bryjolffon and McAffee (2014) have argued that we are now in the second machine age. The first started with James Watt’s steam engine, which kicked off the Industrial Revolution. The second machine age started with using computers and digital tools. Carlopio (1988) has described the different phases of new technology a bit differently. In the 15th and 16th centuries, printing, silk-throwing machinery, the screw press and the windmill were seen as labor-saving devices. Traditional trades were not affected but were actually enhanced. Then the Industrial Revolution began and technology became labor-enslaving, with work processes becoming more standardized and specialized. In the early twentieth century, scientific management became popular, with the assembly line perhaps its most widely used application. Now we are in the third phase of technology which he has described as labor-replacing, where computers and robots will replace jobs.

There has been a lot written on the impact of automation on workers and jobs, and I am not an expert in this area. However, the consensus seems to be that computers are getting smarter. Many experts predict that computers will displace jobs, not just at low-end but also at the high-end. Acemoglu and Autor (2010) suggest that work can be divided into a two-by-two matrix: cognitive versus manual and routine versus nonroutine. Demand for routine tasks to be performed by humans has been falling due to automation, whether these are routine cognitive tasks (e.g., bank tellers, mail clerks) or routine manual tasks (e.g., machine operators, cement masons, dressmakers). But nonroutine cognitive and manual work has been growing, and there has been much debate on the extent to which computers and robots will be able to perform these tasks and replace humans.

In the meantime, we are seeing more and more robots working alongside humans. In 2012, Amazon bought Kiva Systems, a company that makes robots. These robots are used in Amazon’s warehouses to “pick” and bring goods from storage shelves to employees. What do we know about interactions between humans and robots in the work setting? Actually, research on human-robot interaction has been going on for a while. For example, Hinds et al. (2004) wrote about the rise of “professional service” robots (as distinguished from industrial robots) that share the workplace and help people perform their tasks, e.g., supplying troops with ammunition in the battlefield, delivering medications from pharmacies to nursing stations in hospitals. In the future, these robots will be more highly interactive with people. The researchers did a study to determine the effects of the robot’s appearance and its relative status on how people work with robots. They created a lab experiment that required subjects to interact with robots to accomplish some tasks. They manipulated three levels of appearance (human as the baseline; human-like where the robot had a face, torso, arms and legs, and wore an outfit; and machine-like, where the robot covering was metallic and angular). Then they manipulated status by telling the subjects that their robot partner was their supervisor, their peer, or their subordinate. They found that subjects interacting with a more machine-like robot had an increased sense of personal responsibility they felt for the task, with subjects feeling most responsible when interacting with a machine-like subordinate. Furthermore, subjects felt less responsible when collaborating with a robot supervisor as compared with a robot peer or subordinate.

As another example, Kim et al. (2014) reviewed the literature on social distance and developed some hypotheses on people’s reactions to interactions with robots. They had participants play a card-matching game on a computer with Wakamaru, a robot developed by Mitsubishi Heavy Industries, Ltd. As the researchers described it, here is what the robot did:

“In its interactions with the participants, the robot used three key behaviors: gaze, speech, and navigation. The robot’s three-degrees-of-freedom head allowed it to direct its gaze toward the participant and other targets in the environment. The robot communicated with the participants using synthesized natural language and moved toward and away from them at different points during the interaction.” (p. 786)

The participants played the game with the robot, which made suggestions on moves. The researchers manipulated power distance (with the robot either as supervisor or subordinate) and proxemic distance (how close or distant the robot was to the participant).  What they found was that participants who interacted with the supervisor robot at close distances performed better and reported a more positive experience and stronger rapport with the robot than those who interacted with the supervisor robot at far distances. They also found that participants who interacted with a close subordinate robot reported a more positive user experience and more rapport than those who interacted with a distant robot.

What these and other studies show is that there are interaction dynamics between humans and robots that we need to take into account when designing the work of the future. Now if robots can replace humans in many tasks, can they also replace humans as leaders? Here, there is very little research on this topic. I did find an article by Samani et al. (2012) who used the term robotics leadership to describe the work that robots can perform in stock brokering (robots handling stock trades) and avionics (robots replacing human pilots in airplanes). However, I don’t consider these as leader behaviors as much as expertise that robots can provide. There is also some emerging research on leader-like behaviors of robots among themselves; for example, using simulation techniques among foraging robots, Pugliese et al. (2015) found that the most skilled robots became “leaders” and that robot groups with leaders were more effective than robot groups without leaders.

Part of being a leader is influencing others to perform certain actions. We have seen evidence from research by Milgram (1963) and others on the effect that those perceived to be in positions of authority can have on getting others to comply and follow orders. Parasuraman et al. (1997) wrote about this in the context of automation, when they summarized the research on the dangers of automation. A robot leader who gives instructions (especially if the robot has a deep male voice and perhaps is made up to look dominant) might have others following orders unquestioningly. In fact, in an interesting experiment where participants worked with robots on a set of tasks, Gombolay et al. (2015) found that participants preferred robots who made decisions about how the tasks were to be allocated rather than the participants having to make the decisions themselves. In another experiment, Robinette et al. (2015) found that participants followed the robot in an artificially created emergency situation (where they had to be evacuated from a room) even when they had seen the robot make navigation mistakes earlier and continued to make mistakes in directing them to a wrong exit!   

We do know quite a bit about effective leadership, and what people look for in leaders. Kousez and Posner (2007) have been conducting surveys over the past 30 years on qualities most admired in leaders, and they have found that four qualities are what most of the people they surveyed around the world want in a leader: honest, forward-looing, inspiring, and competent.

In his now classic article, Kotter (2001) stated that leaders do things differently than managers: leaders set a direction, align people, and motivate them. More recently, Google found in its own research (Garvin, 2013) that outstanding managers (versus average managers): coach well, empower their teams and do not micro-manage, express interest in employees’ success and well-being, are productive and results-oriented, are good communicators and listen to their teams, help their employees with career development, have a clear vision and strategy for the team, and have key technical skills so they can help advise the team.
Can robots be as good as humans in performing these behaviors? Potentially, yes. I can envision robot leaders being programmed, for example, to express interest in employees’ success, to listen and provide a clear vision for the team. I can also envision robot leaders being programmed to be honest, forward-looking, inspiring, and competent.

However, effective leadership is not only about performing these behaviors. As Kousez and Posner (2007) argue, leadership is about establishing a relationship between leader and follower. While certain robot leadership behaviors might lead to compliance among followers, organizations also want to create high-performing cultures characterized by motivation and commitment. Can robots get us to trust them so that they inspire us to do our best and engage us? Can a robot ever cause our brains to release oxytocin, which is a chemical that helps promote many kinds of social behavior (Stix, 2014)? Based on the evidence, it seems that our brains are wired differently when it comes to reacting to those who inspire, engage, and motivate us versus those who simply get us to comply.

As Colvin (2015) has pointed out, we are asking the wrong question if we are trying to figure out only what computers cannot do that humans can: “Rather than ask what computers can’t do, it’s much more useful to ask what people are compelled to do – those things that a million years of evolution cause us to value and seek from other humans, maybe for a good reason, maybe for no good reason, but it’s the way we are.” (p. 53)

In my view, we we will continue to want human leaders in the work setting especially in three (and very “human”) areas where leadership is critical: making decisions on business and people issues (specifically around strategic decisions, and on who to select and promote), communicating those decisions and related issues, and inspiring and motivating. The value-add of human leaders is evident in these areas, where commitment rather than compliance is of critical importance to building a high-performance organization.

For each of the above, note the following continuing patterns:
·       Despite years of evidence that statistical methods of selecting job candidates are superior to human methods, almost every firm that I know still wants to see a candidate face-to-face before hiring him or her, especially for higher-end work and/or professional positions.
·       When managers congratulate someone for a job well done, or let them know that they have been promoted, they much prefer to do this face-to-face rather than sending them an e-mail or handing them a piece of paper with the news. The same is true for communicating bad news, such as when someone has to be let go. Evidence suggests that people also would much prefer to hear this information face-to-face from their leaders.
·       When a team needs to be inspired and motivated, the most effective managers engage in face-to-face meetings and “high-touch” actions to lift spirits up and boost morale.

Brynjolfsson, E. and McAfee, A. (2014). The Second Machine Age: Work, Progress, and Prosperity in a Time of Brilliant Technologies. New York: W. W. Norton.

Carlopio, James. (1998). Implementation: Making Workplace Innovation and Technical Change Happen. Synergy Books International.

Colvin, G. (2015). Humans Are Underrated: What High Achievers Know That Brilliant Machines Never Will. New York: Penguin.

The Economist (2016). Who Wields the Knife? May 17, p. 74.

Garvin, D. (2013). How Google Sold Its Engineers on Management. Harvard Business Review.

Gombolay, Matthew C., et al. (2015). Decision-making authority, team efficiency and human worker satisfaction in mixed human–robot teams. Autonomous Robots 39 (3): 293-312.

Hinds, P., Roberts, R. and Jones, H. (2004). Whose Job Is It Anyway? A Study of Human-Robot Interaction in a Collaborative Task. Human-Computer Interaction, 19: 151-181.

Kim, Y. and Mutlu, B. (2014). How Social Distance Shapes Human-Robot Interaction. International Journal of Human-Computer Studies, 72: 783-795.

Kotter, J. (2001). What Leaders Really Do. Harvard Business Review.

Kousez, J. and Posner, B. (2007). The Leadership Challenge (Fourth Edition). New York: Wiley.

Milgram, S. (1963). Behavioral study of obedience. The Journal of Abnormal and Social Psychology67 (4), 371-378.

Parasuruman, R. and Riley, V. (1997). Humans and Automation: Use, Misuse, Disuse, and Abuse. Human Factors, 39 (2): 230-253.

Pugliese, F. et al. (2015). Emergence of Leaders in a Group of Autonomous Robots. PLoS ONE 10(9): e0137234. doi:10.1371/journal.pone.0137234.

Robinette, P et al. (2015). Overtrust of Robots in Emergency Evacuation Scenarios.

Samani, H. et al. (2012). Towards Robotic Leadership. In SS. Par et al., (Eds.), ACHRS Part II. Hedelberg: Springer, pp. 158-165.

Stix, G. (2014). Fact or Fiction: Oxytocin Is the Love Hormone. Scientific American, September 8: