Notes

Addressing Bias and Promoting Diversity

Identifying the Source of Bias

When organizations face issues related to bias, it's crucial to pinpoint the source to implement effective strategies. The discussion covers how to find the source of bias, develop strategies, measure impact, identify key metrics, and go beyond mere correlation to establish causation.

Correlation vs. Causation

Experimental Work

Establishing causation requires experimental work, which can be challenging in organizational settings.

Pre-Post Test

An alternative approach is a pre-post test, which gets closer to causation but doesn't fully establish it.

Importance of Understanding Data Collection

It's essential to deeply understand data collection methods and the conclusions that can be drawn from them. In today's environment, data is readily available, but conclusions drawn from it may not always be accurate.

Challenges in Diversity

Different Perspectives

The answers to questions about the challenges to diversity vary depending on who you ask because people often depend on intuition and personal experiences.

Gender Diversity Example

• Men's Perspective: Men often see recruitment as the biggest obstacle to gender diversity.

• Women's Perspective: Women often view advancement and promotion pathways as the primary challenge.

The Need to Look at Data

Instead of relying on personal opinions, it's crucial to collect and analyze data to understand the actual issues.

Addressing Gender Disparities in Promotions

Case Scenario

A consulting firm hired to address gender disparities in promotions, where hiring is evenly split but a gap emerges at the leadership level.

Identifying the Source

The key question is, "What is the source of the problem, and how can we get to it?" Promotions should not be seen as a single event but as a process involving mentorship, assignments, and performance reviews.

Targeting Strategy

Targeting the source of the problem is most effective. Addressing the source minimizes the need for extensive work at other levels. However, if accessing the source is difficult, multiple strategies may be necessary.

Intervention Design

When designing interventions, it's important to match the intervention to the specific issue identified through data.

Yelp Case Study: Hiring Disparities

The Problem

Yelp faced ongoing disparities in hiring, particularly with college recruits. They targeted this specific area for intervention.

Hiring as a Process

Hiring is a process, not a single event. Yelp broke down their process to identify issues at each stage

Stages of Hiring

1. Potential Applicants Decide to Apply: Job ads and the schools where they were placed were examined.

2. Decision to Interview: Resume screening and coding tests were analyzed.

3. Interviewers Evaluate Applicants: Skype and on-site interviews were reviewed.

4. Yelp Makes Hiring Decision.

5. Applicants Make Decision on Offers.

Screening Resumes

Yelp's biggest issue was the strict criteria used during resume screening. They were primarily focused on specific STEM majors, overlooking talented individuals with other backgrounds, such as history or psychology.

Arbitrary Rules

The arbitrary rule of only considering specific majors led to missed opportunities with talented individuals who had relevant experience, such as founding coding camps or interning at tech companies.

Impact of Screening

This screening process disproportionately affected women who might have coding skills but didn't pursue traditional STEM majors due to the male-dominated environment.

Strategic Consulting

When consulting on diversity, identify where the lack of diversity begins or increases. Avoid assuming that best practices will universally fix everything, as each organization is unique.

Data Collection

Collect demographic data during job applications to assess the applicant pool's composition. For example, if the applicant pool is 60% men and 40% women, this serves as a baseline.

Analyzing Data

If the interview invitation process results in 80% men and only 20% women, it signals potential bias at this stage. Investigate the screening criteria, coding tests, and resume scoring methods.

Addressing Issues at Every Stage

Yelp found issues at every stage and implemented multiple strategies to address them.

Updating Campus Visits

Yelp updated the campuses they visited, expanding their reach to include more UCs and CSUs, and eventually community colleges. They found qualified candidates at schools beyond their initial target list.

Setting Goals

Yelp set a goal for new sales hires to match the demographics of the city they operated in, as it improved client connections. It's important to distinguish between a target goal and an illegal quota.

Avoiding Quotas

Quotas, which are mandatory hiring targets, are illegal. A target goal, on the other hand, provides information about progress toward parity. If the goal isn't met, the process should be re-evaluated to identify missed opportunities or biases.

Yelp's Actions

• Added 10 new schools to their recruiting list to increase gender diversity.

• Hired a dedicated recruiter to ensure a broad reach.

• Sent a female engineer to represent Yelp during campus visits, which increased interest from both men and women.

• Removed the initial resume screen to avoid screening out candidates based on their major.

• Introduced training for interviewers, as they previously conducted unstructured interviews.

• Experimented with blinding resumes and altering voices during interviews, but this didn't prove helpful.

Results

By 2017, these strategies increased women's presence in tech positions from 10% to 18%, marking an 80% change in about three years.

Presenting Impact

It's way more impactful to present the improvements using percentages. For instance, emphasize the 80% increase in women in tech positions rather than just presenting the 18% figure.

Developing a Strategy

Make decisions like a scientist by:

• Developing hypotheses about the source of the problem.

• Using data to identify disparities.

• Picking research-based strategies tailored to the context.

Context Matters

Context of the organization matters a lot which includes the organization's current state and culture. For example, acknowledging the value of diversity may only be effective if the organization already reflects diversity.

Strategies

• Acknowledging the Value of Diversity: Only effective if genuinely reflected in the organization.

• Increasing Intergroup Interactions: Not helpful if the organization lacks diversity.

• Applying Consistent Criteria for Hiring: Only effective if the criteria is unbiased.

• Rotating Office Housework Tasks: May not work if there's an imbalance in representation.

Measuring Impact

Evaluate the impact of diversity initiatives, as what is not measured cannot be fixed.

Key Metrics

• Metrics of Success: Short-term and long-term goals.

• Assessing Progress:

  • Measuring parity in ratings in performance reviews

  • Assessing the actual number of promotions over different durations.

  • Promotion strategies are relatively easy to measure success using numbers and breakdowns compared to inclusion strategies.

Measuring Inclusion

Measuring inclusion requires surveys to assess feelings of belonging and uniqueness. Avoid vague questions.

Improving Survey Questions

Instead of asking vague questions like, "How much belonging do you feel?" specific questions about behaviors are needed. For example, "How often does your manager or coworker seek out different viewpoints?"

Specificity in Questions

Specificity helps in targeting interventions, like managerial training for leaders who don't seek different point of views and team building where co-workers don't.

Evaluating Progress

Consider the Starbucks racial bias training example. It is important to check that the diversity training was effective in increasing desire to foster inclusion and measure real impacts.

Causal Inference

Correlation vs. Causation Reminder

Remember that correlation does not equal causation.

Considering Factors

Attending bias training leads to desire to foster inclusion, which means that training increases desire to show diversity.

Attending Bias Training

People tend to go in person if they are more likely to attend even if they have desire to foster inclusion.

Critical Thinking

Apply critical thinking skills to data presented, as it can be manipulated. Here's an example: People drank wine at a party and were more healthy: This may be a socioeconomic connection because of more healthcare.

Alternative Explanations

Consider alternative explanations when analyzing data. Example: Correlation between ice cream sales and homicides. A third factor is summer: the heat makes people angrier and leads to ice cream since it is hot.

Reverse Causation

It is important to recognize that correlations do not mean cause and effect. You can instead say that there is a relationship, an association, or correlation.

Experimental Studies

To truly measure the effect of diversity training, here are variables to consider:

• Have people take a randomized test and control group.

• Use a pre-post-test, in which test outcomes show test scores and results before implementation.

Experiments to Establish Causation

Important when designing an experiment: You have to assign employees to training with or without other participation. The standard that you will be measured to is the gold one to help establish this causation.

Running Experiments

1. Randomly assign participants

2. Independent variable

3. Dependent variable which is outcome you change by changing it

Pre Post Tests

Measure at Time 1 and during implementation to allow the offer to show up and show how to have the best training possible.