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 15^th and 16^th 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 Psychology, 67 (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. http://www.cc.gatech.edu/~alanwags/pubs/Robinette-HRI-2016.pdf

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: http://www.scientificamerican.com/article/fact-or-fiction-oxytocin-is-the-love-hormone/

Is Meritocracy Dying, Or Was It Really Just a Myth All Along?

An organizational culture that values meritocracy sounds good, doesn’t it? After all, we want to be hired, evaluated, and promoted based on our own merits – as opposed to non-performance-related factors such as our race, our gender, our network, or our political skills. It’s all about getting what we deserve because of our performance.

Scan the web sites of companies and many of them make explicit reference to meritocracy. Goldman Sachs states it is “a meritocracy built on the belief that collaboration, teamwork and integrity create the right environment for our people to deliver the best possible results for our clients.” One of McKinsey & Company’s core values is to “sustain a caring meritocracy.” Similarly, one of Bank of America’s core values is “inclusive meritocracy.” AB InBev, the beer conglomerate, claims that its culture “is built on ownership, informality, candor, transparency and meritocracy. We set ourselves stretch targets and are never completely satisfied with our results.” The newly formed Kraft Heinz Company states on its web site that “we recognize and reward outstanding performance at every level, in the true spirit of meritocracy. In 2014 alone, more than 1,000 employees were promoted as a result of their high performance and value creation.” And there is GE, which has meritocracy as one of its values, which they define as “creating opportunities for the best people from around the world to grow and live their dreams.”

Not only employers but most employees also seem to value meritocracy. According to Ready et al. (2010), this is one of the factors that attracts talent in emerging markets. In my experience, I have also found that young professionals in Asia find the seniority-based system in many Asian firms to be hampering their advancement opportunities, and they want more recognition for their hard work and performance. In fact, when China recruited for its civil service examinations starting around the seventh century, it introduced a merit-based system. The exam was based on classical literature and philosophy, and candidates were selected based on their exam scores. Of course, there were many people from the lower classes who did not have the educational background to compete, and women were excluded from applying. In principle, however, the concept was based on a meritocracy.

Yet there are significant barriers to creating meritocracy in organizations. First, organizations may say they value it but their practices might not necessarily reflect this. Executive coach Marshall Goldsmith (2004) has written about how leaders subtly encourage (and by implication, reward) those who fawn over and suck up to them. According to a recent survey by Right Associates (Lauby, 2012), 44% of employees believe the key to their success lies in “who you know,” and not job performance (39%). Other employee surveys show similar results, despite the prevalence of organizational practices such as performance appraisal systems.  In a 2014 survey of over 356,000 U.S. federal employees, 54% disagreed with the statement that “pay raises depend on how well employees perform their job.” (fedrev report, 2014) In their excellent book, The Meritocracy Myth, McNamee and Miller (2004) point to social capital (“whom you know”) and cultural capital (“what you need to know to fit into the group”) as important non-merit factors that influence life outcomes:

“It helps to have friends in high places, and the higher up one starts in life the greater the probability that one will travel in high-powered social circles. One must also have the cultural wherewithal to be fully accepted within these high-powered social circles. Those who are born into these circles have a nonmerit cultural advantage over those not born into these circles …” (p. 198)

Second, as Castilla and Benard (2010) have argued, emphasizing meritocracy might unintentionally introduce bias and create inequity in the distribution of employee rewards. Their research suggests that “… in contexts in which people are led to feel that they are unbiased, fair, or objective, they are more likely to then behave in biased ways.” (p. 547); they refer to this as the “paradox of meritocracy effect.” Why does this happen? Two mechanisms seem to be operating. One is through the concept of moral credentials. If you believe you have established your credentials as a non-prejudiced person, you are more prone to express prejudiced attitudes.  Similarly, if you believe strongly that your organization values and endorses meritocracy, then you must therefore not be biased and must be making your decisions based solely on merit. The second mechanism is through self-perceived objectivity. If you believe that you are an objective person, then you may not realize that you may be acting on your prejudices. And we all have conscious and unconscious biases. So the paradox here is that if you believe in meritocracy you might be more likely to behave in biased ways.

Third, we are subject to the fundamental attribution error, which is the tendency to overattribute other people’s behavior to internal rather than to external causes, and one’s own success to internal rather than external causes. Those who have made it and who are successful believe that it is due to their hard work and because they got there through their merits. They also believe that internal factors such as aptitude or motivation are the causes for others not succeeding, rather than external circumstances.

A fourth barrier is around our unconscious biases, especially related to (but not limited to) race and gender. As an example, Samson (2013) conducted a survey-based experiment in which half of the sample of White respondents in California were asked about the importance of GPA (a merit-based variable) as a factor for admission to the University of California. The other half received the following information prior to the question: Under current admissions procedures in the University of California system, Asians make up almost 40% of the student body (or 2 out of every 5 students), while they are only 12% of the California population (p. 240). His results showed that individuals’ acceptance of merit-based admission varies as a function of their perception of “group threat” and not just perceptions of fairness. In other words, when group threat is primed (as it was in half the subjects), these White subjects tended to decrease the importance of a merit-based variable like GPA as a factor in admissions. He concludes that commitment to meritocracy may be based not solely on principle but varies “… depending on the outgroups under consideration and the extent of the group threat they pose to Whites.” (p. 253)  

In an interesting set of studies, Uhlmann and Cohen (2005) showed that individuals use different criteria to define merit that fits their own preconceptions and biases. For example, in one study, they asked participants to rate the strengths of male and female applicants for the job of police chief. Applicants’ credentials and gender were manipulated. Credentials were either strong on being streetwise (e.g., working in tough neighborhoods, getting along with fellow officers, poorly educated) or on being educated (e.g., experienced in administrated, well educated, did not get along so well with fellow officers). The participants then rated the importance of several hiring criteria, and whether the applicant should be hired. What they found was that educated characteristics were more important when the male applicant had them than when he did not. Educated characteristics were less important when the female applicant had them. In fact, male and female participants constructed criteria favorable to the male applicant. Because the job of a police chief is traditionally male, the participants defined merit in a way that favored male over female applicants.    

As reported in the Wall Street Journal (Silverman, 2015), the Clayman Institute for Gender Research at Stanford has been analyzing hundreds of performance reviews in several technology companies. So far, researchers have found that women received 2.5 times the amount of feedback men did about aggressive communication styles (e.g., suggesting that they pipe down and be less aggressive), while men’s reviews contained twice as many words related to assertiveness, independence, and self-confidence (e.g., “drive,” “transform,” “tackle”). In addition:

“… women’s reviews had more than twice the references to team accomplishments, rather than individual achievements … while men … received three times as much feedback linked to a specific business outcome and twice the number of references to their technical expertise.”

Fifth, the criteria for determining merit can vary by company and by culture. Being promoted in some Asian companies means that you have to be able to speak English very fairly well. Many companies look at factors other than performance to determine who gets hired, rewarded and promoted, and arguably some of these factors are more subjective than others, as well as being weighted differently by different raters. For example, in identifying those who might be the future leaders of the company, executives might consider performance, but often look at other factors, such as interpersonal skills, that elusive quality called executive presence, and perhaps even the advocacy of other executives who might have mentored these individuals. Aren’t some of these “non-merit” factors? McNamee and Miller (2004) give a detailed example of their own experience selecting a new faculty member for their department, and they conclude:

“… we call into question the presumption that people know merit when they see it. … it is a cardinal principle in meritocracy that the ‘most’ qualified or ‘best’ person should be hired for the job … However, we argue that it is often difficult or impossible to ‘know’ who the ‘best’ is.” (p. 43)

What can organizations do? First, and the most basic, is to make sure their practices are consistent with their statements about merit. Do their performance and reward systems clearly reinforce excellence and high performance? They can conduct periodic surveys or pulse checks to determine what employees’ perceptions are, instead of simply relying on making pronouncements. Second, make these practices and processes more transparent by defining more clearly what excellent performance is, and why it matters. For example, if behaviors in addition to performance are used as criteria for rewards and promotions, make these explicit, and build in safeguards to minimize subjectivity.

Third, hold managers accountable. Have processes in place to make sure that managers evaluate and reward employees based primarily on performance and other merit-based factors. There is a lot of research evidence from studies of small groups that meritocratic hierarchies (for example, when individuals in groups are given higher rank because of their task expertise) predict group success (Anderson and Brown, 2010).

Fourth, raise awareness of unconscious biases. According to the Wall Street Journal (Lublin, 2014), as many as 20% of large corporations with diversity programs (including Chubb, Genentech, Google, Price Waterhouse Coopers, Roche and T. Rowe Price) now provide training on unconscious bias. Other companies (e.g., Microsoft) provide guidelines to managers before they write up their performance reviews to remind them about gender bias.  There is some evidence that such training might be helpful. Of course efforts to consider diversity in hiring and promotion decisions should continue. Especially important is making sure that there is diversity in the pool of candidates being considered. Most White senior executives when considering successors for their positions or those within their team might favor first those who they perceive to be like them (the “similar-to-me” bias). With a more rigorous screening process, organizations can make sure there are candidates who may be just as qualified but who may not be on their radar screen – and some of them might be female and people of color.

Fifth, as much as possible, define criteria for merit up front and get agreement from evaluators on what the indicators for these criteria are. This will reduce individuals’ tendencies to emphasize qualifications that tend to fit their biases. When working with organizations on evaluating their future leaders, make it a practice, as my colleagues and I do, to gain agreement with the executive team on what success looks like in concrete terms.

It is clearly not enough for organizations to simply state that they have a meritocratic culture. In fact, the reality is that it is very difficult for any organization to achieve this ideal, although it is well worth the effort. The current debates about wage and income inequality touch on the difficulties of achieving meritocracy in societies. However, companies that are serious in attracting and retaining talented employees need to continue to make serious efforts to eliminate these barriers.

Anderson, C. and Brown, C. (2010). The Functions and Dysfunctions of Hierarchy. In A. P. Brief and B. Staw (Eds.), Research in Organizational Behavior, 30: 55-89. New York: Elsevier.

Castilla, E. and Benard, S. (2010). The Paradox of Meritocracy in Organizations. Administrative Science Quarterly, 55: 543-576.

Goldsmith, M. (2004). http://www.marshallgoldsmithlibrary.com/cim/articles_display.php?aid=102

Kaplan, S. (2015). Meritocracy: From Myth to Reality. Rotman Magazine, Spring, 49-53.

Lublin, J. (2014). Bringing Hidden Biases into the Light. Wall Street Journal, January 9:  http://www.wsj.com/articles/SB10001424052702303754404579308562690896896

Lauby, S. (2012). For Your Career, It’s Not What You Know – It’s Who You Know. http://blog.shrm.org/blog/for-your-career-its-not-what-you-know-its-who-you-know

Samson, F. (2013). Multiple Group Threat and Malleable White Attitudes Towards Academic Merit. Du Bois Review, 10(1): 233-360.

https://www.fedview.opm.gov/2014/Reports/ResponsesW.asp?AGY=ALL

Silverman, R. (2015). Managers: Watch Your Language. Wall Street Journal, September 30 2015.

Uhlmann, E. and Cohen, G. (2005). Constructed Criteria: Redefining Merit to Justify Discrimination. Psychological Science, 16 (6): 474-480.