Removing Workplace Biases with Behavioural Design
Q&A with Iris Bohnet, Harvard Kennedy School Professor
Interview by Karen Christensen
There is some disagreement about the ‘business case’ for gender equality. What is your take on it?
The disagreement is justified. The focus to date has largely been on the diversity of corporate boards and senior management teams, and the problem is, we don’t have the data required to make solid conclusions. Even when we find a correlation between gender diversity on a board and a company’s performance, we have no way of proving that there is a causal relationship going on.
Recently, a meta-analysis came out, summarizing about 120 studies, and it did find a small positive correlation between gender diversity and overall firm performance. But again, this was a correlation, not causation. If we want to establish causality, we will have to create teams randomly and measure whether the more diverse teams outperform the homogeneous teams. Some of the best work in this area has been done in the realm of ‘collective intelligence’ (i.e. the intelligence of groups). This research has found a strong causal relationship between gender diversity and team performance across many different tasks.
As a result, I believe we have enough evidence at the micro level that a business case exists. However, I’d love to see us move this discussion beyond a numbers game, and start to think more about fostering inclusive behaviour.
How do you define ‘behavioural design’?
The research shows that we can’t help but put people into categories, and behavioural design builds upon this element of how our minds work. Basically, it uses behavioural insights to de-bias organizational practices and procedures, rather than focusing on changing mindsets. Within an individual mind, biases tend to occur automatically and unconsciously, and it’s really hard to change that. It’s much easier to take steps to de-bias an organization.
I’d love to see us move this discussion beyond a numbers game, and start to think more about fostering inclusive behaviour.
Do diversity training programs work?
We don’t really know, because most organizations don’t measure the results — and the few that do have generally found that they don’t work. We have some correlational data looking at whether or not a company has a diversity training program and the actual diversity of its workforce, and in short, that correlation does not exist. So the picture is not optimistic.
A few companies are trying innovative approaches — from implicit bias training to programs aimed at specific inequalities. Carnegie Mellon’s Linda Babcock and George Loewenstein have researched the effectiveness of various de-biasing techniques. One intervention they studied is ‘perspective taking’, which simply means trying to walk in your counterpart’s shoes, take their perspective and understand where they are coming from. For example, ‘walking in an elderly person’s shoes’ by writing an essay from their perspective was shown to reduce stereotypes about the elderly.
Babcock and Loewenstein also experimented with a ‘consider the opposite’ strategy, which involves being your own devil’s advocate and questioning your assumptions — actually coming up with arguments for why your thinking might be wrong. This has been shown to work — but it requires a lot of maturity and self-awareness to be able to question yourself. It’s easier if someone else does the ‘heavy lifting’ for you.
Given all the evidence, I would urge companies to focus their training programs on capacity building and adopt the ‘unfreeze-change-refreeze’ framework — a method borrowed from my Harvard colleague, Max Bazerman. Successful ‘unfreezing’ happens when people start to question their current strategies and become curious about alternatives. Once ‘unfrozen’, you spend some time on what your organization is currently doing, and what could change. Finally, you think of ways to ‘refreeze’ the new insights gained and the new behaviours learned. In the end, the pathway to behavioural change may not be a change in individual beliefs but instead a change in socially-shared definitions of ‘appropriate behaviour’.
One of the more recent applications of Big Data in the workplace is ‘people analytics’. Please describe how it works.
This basically entails bringing the rigour of your finance or marketing department to HR, arguing that data can help us better predict, for example, the future performance of a particular job candidate than the best interview ever could. It involves moving away from intuition and building on data.
The question is, What kind of data? Organizations can use all sorts of data points, but one powerful example is ‘looking backwards’: You can use data and machine learning to basically learn from the past. For example, you could take a close look at the data points for ‘individuals who have been highly successful’ in your organization: What are their shared characteristics? You might look at which universities they went to, and find that it’s a good thing not to come from an Ivy League school — or maybe that it’s better to have an Engineering background than a Math background.
Many industries still suffer from a ‘leaky pipeline’—a metaphor for the continuous loss of women as they climb the career ladder. What can be done to address this?
The leaky pipeline argues that we don’t have, for example, enough female engineers, or engineers with racially-diverse backgrounds, and that is still true. One critical use of data is to use it to understand what is broken in your organization. For some — for example, law firms — the leaky pipeline is no longer an issue in terms of gender. We now have more than 50 per cent women graduating from law schools — as well as an increasing number of people of colour, so for law firms, we don’t have to talk about the leaky pipeline anymore. It’s more an issue of progression within each particular firm.
However, in sectors such as science, engineering and technology — the STEM fields — women and people of colour are still under-represented, so the leaky pipeline persists. We need to go into our schools and universities early on, and start to nourish interest in these fields. That might include de-biasing our classrooms, de-biasing the way we teach, and providing counter-stereotypical role models. For example, bringing in a female Math teacher or engineer to speak to a class — as well as male nurses or male English teachers. For both boys and girls, these experiences can be extremely powerful.
Google’s HR department has been referred to as an ‘employee science lab’. Describe some of its innovative practices.
In terms of using data to understand what is broken at the company, Google has done great work, and it has also made progress in terms of fixing what is broken. One example relates to gender. By doing a data analysis, Google realized that women were more likely to leave the firm or quit. When they dug a bit deeper, they noticed that most of these employees were young mothers, and this enabled them to design an intervention targeting this group: They dramatically increased parental leave, both for mothers and fathers.
They also found some really interesting patterns in terms of their hiring practices. Obviously, Google has grown dramatically in recent years, and it has had to adjust to that enormous growth. Initially, it would have groups of current employees interview new job candidates. But when it did some Big Data analytics, it found that the magical number of interviewers was actually just four people — at which point, the rankings started to converge, and a fifth person didn’t add much value. Recently, they published some fascinating work around, What makes an employee perform well? They found that the biggest factor was psychological safety within teams.
What are the initial steps involved in ‘designing diversity’?
Instead of just throwing money at the problem, you should start by using data to understand what is broken. Once you understand whether the problem is a leaky pipeline or progression within your organization, you can start to design and test interventions. I’m hopeful that organizations will learn from their marketing departments and do more ‘pilots’, A/B testing and other experimental approaches, to try things out and measure the results.
Another thing you can do is move from unstructured interviews to structured interviews, and see what happens, or de-bias the language in your job postings. Maybe you could take a traditional version of a job posting and a proactively de-biased version, examine what difference this makes to the type of people you attract, and learn from that.
We need to go into our schools and universities early on, and start to nourish interest in the STEM fields.
If your problem is not attracting talent, but progression within your company — which is still is a big issue for most organizations — you might want to look at your performance appraisals and promotion procedures and de-bias them, as well. One idea is the practice of sharing self-evaluations with managers before the manager makes up his or her mind in a performance appraisal. Of course, behavioural insights tell us that this will create an ‘anchor’ and influence the manager; and when people differ in their self-confidence or their willingness to brag, they will evaluate themselves either less or more harshly, and that will translate to how managers see them. In this regard, there are differences not just across genders, but also across cultures — so there’s a lot to look out for.
Looking ahead, are you optimistic that firms will become more inclusive?
The good news is that a number of start-ups have embraced behavioural insights and translated them into software, making it much easier for organizations to do this kind of work. Applied, for example, is a company that focuses on de-biasing a company’s hiring procedures. For example, blinding hiring managers to demographic characteristics, and designing work sample tests and structured interviews. They also have software that can debias the language in job advertisements and evaluate the impact. Another company is Pymetrics, which was started by a team of neuroscientists and MBAs from MIT and Harvard. They are using gaming tools to help people assess themselves and what types of jobs might make them happy.
All of this work makes me optimistic about the future of behavioural design. The bottom line is that bias is built into our practices and procedures, not just into our minds, and bad designs — whether consciously or unconsciously chosen — lead to bad outcomes. Through behavioural design, we can change behaviour by changing environments rather than mindsets.
Behavioural Economist Iris Bohnet
is a Professor of Public Policy at the Harvard Kennedy School, director of its Women and Public Policy Program, co-chair of its Behavioural Insights Group, an associate director of the Harvard Decision Science Laboratory, and the faculty chair of the executive program “Global Leadership and Public Policy for the 21st Century” for the World Economic Forum’s Young Global Leaders. She is the author of What Works: Gender Equality by Design
, a 2016 Financial Times
Best Business Book of the Year.
This article originally appeared in The Inequality Issue (Fall 2017) of Rotman Management. The magazine offers the latest thinking on leadership and innovation and is published three times a year.
Share this article: