Carra Sims is a senior behavioral and social scientist at RAND. She's also a member of the Society for Industrial and Organizational Psychology, which recently asked her about one of its top ten workplace issues to watch in 2019: sexual harassment and what's next for the #MeToo movement.
As a researcher who has studied sexual harassment in the workplace for over a decade, what is your perspective on the changes that have taken place in recent years?
Interest in the topic of harassment seems to come in waves, with each wave getting us a little farther up the beach. If you go back, in the early '90s you had the Tailhook scandal and Anita Hill really helping to move the discussion into the national consciousness. Then people seemed to move on. A few years later, the scandal at Aberdeen Proving Ground broke, and you had the harassment class-action lawsuits on Wall Street. Now, we have #MeToo.
Today, two things are different. First, there is more conversation around the daily low-grade behaviors that contribute to environments in which harassing behaviors are seen as acceptable. And there's more conversation around how you have to pay attention to these low-grade behaviors as well, stopping negative tendencies before they develop. Second, there has been a broadening of the conversation to include industries that often get less focus, such as fast food service.
So, I do think things are getting better. My perspective is that we are moving forward but that the underlying societal cause of sexual harassment—that women are still not quite seen as equal—remains. We're making incremental improvement on average: We move a bit forward and then get a reminder that the problem isn't solved. We move a bit more forward, and there's another reminder, or backlash. But the overall trend is positive.
Recently, 181 CEOs from major U.S. corporations, endorsed the idea that corporations should focus on all stakeholders (which includes employees), not just shareholders. How can this type of change in organizational objectives be leveraged to prevent sexual harassment in the workplace?
As a researcher, I emphasize the importance of looking at actual behavior. For harassment itself, the issue is that people can experience behaviors that meet legal criteria of harassment, but they may not label it as harassment, so it's very important to make sure you center your definition in the behaviors people are experiencing or engaging in. I would apply this to the statement from the Business Roundtable about the focus on all stakeholders—I want to see how this gets put into practice, behaviorally.
Maximization of shareholder value has been associated with a shorter-term perspective—doing things like stock buybacks rather than capital investment—and increasing inequality via compensating executives with stock options. Looking to maximize employee well-being as well as shareholder profits could engender a longer-term focus and a focus on improvements like enabling employees to share in the gains from productivity. On the harassment front, being more respectful of all employees could engender less tolerance of things like toxic leadership behaviors, where supervisors run their people into the ground in order to get short-term results and such abusive behavior is rewarded. Obviously, sexual harassment is itself very detrimental to the well-being of employees who experience it, and so efforts to enhance employee well-being could include things like an emphasis on civility and respect in the workplace for everyone, which would help set norms that are less tolerant of sexual harassment, specifically.
Given the rapid rate of change in the way many organizations are functioning (e.g., matrixed, gig economy), what associated challenges are organizations facing in terms of sexual harassment?
I'll go back to the issue of norms. For organizational incentives to be aligned to produce an environment in which sexual harassment is not tolerated, there has to be accountability. If being in a matrixed organization means that it's harder to hold employees who engage in harassment accountable, because reporting lines are confused or because authority to impose appropriate sanctions is diffuse, that's a problem.
The best way to determine prevalence of sexual harassment is through surveys, and due to liability concerns organizations just don't survey employees about such issues.Share on Twitter
The issue with the gig economy is a bit different. Organizations like Uber and Lyft claim that drivers are not actually their employees, and they are not liable for driver behavior, because drivers are independent contractors. So the concern is that no one is getting held accountable. Vetting of drivers or other gig economy employees is not getting done appropriately and with a transient workforce, harassment behaviors may or may not be reliably documented and follow a perpetrator.
Bad behavior happens when people think they can get away with it, so to the extent that things like matrixed organizations, the gig economy, and offloading work to contractors rather than permanent employees contributes to a weak and ambiguous situation, individuals' inclinations to behave badly are offered more free reign. Detecting habitual offenders and holding them accountable can be much harder.
Weaving together additional topics from SIOP's top 10 list, what role will machine learning and its associated big data play in how organizations understand and deal with sexual harassment?
While there is some promise here, sexual harassment is a challenging domain for application of these tools for a number of reasons. Machine learning offers a promise of procedural consistency which would definitely be beneficial in managing sexual harassment. From an employee perspective right now, treatment of harassers can seem pretty arbitrary and opaque (due to confidentiality clauses that can result in a lack of visibility over the scope and outcome of a complaint for both accused and accusers).
However, algorithms are only probabilistically correct, and that's when all the right assumptions are made and they are based on perfect data. We definitely don't have perfect data when it comes to harassment, either in terms of actual incidence or reporting or contextual factors.
The best way to determine prevalence of sexual harassment is through surveys, and due to liability concerns organizations just don't survey employees about such issues. An exception is the military, where it's required by law. Without survey data, organizations would be reliant on personnel complaints to develop algorithms, a process that has known flaws. Even employees who experience behavior that qualifies are not necessarily likely to label it as harassment, and even fewer go on to report it due to legitimate concerns about retaliation. So the data on which any machine learning algorithms would be trained is skewed in a way that benefits perpetrators of harassment rather than employees experiencing harassment. Incautious use of algorithms has been found to further institutionalize discrimination in credit scoring and estimation of criminal risk, for example—this would presumably also be the case for harassment.
We'd need to make further progress on how to treat ethical issues in machine learning, and be comfortable with the collection of pretty intrusive levels of data, to really be able to use big data and machine learning to alleviate sexual harassment incidence.
Implementing these tools to combat harassment becomes even more challenging when you add in the necessity of transparency about the data, assumptions, and implementation—which is not necessarily customary now for machine learning—and the fact that some organizations may really minimize the information available about complaints due to confidentiality concerns.
Carra S. Sims is a senior behavioral and social scientist at the nonprofit, nonpartisan RAND Corporation.
This commentary originally appeared on Society for Industrial and Organizational Psychology on September 18, 2019. Commentary gives RAND researchers a platform to convey insights based on their professional expertise and often on their peer-reviewed research and analysis.