Insights
Human-Centered AI for Human Resources: A Q&A with The Conference Board's Stela Lupushor
Organizations across the world are increasingly interested in using artificial intelligence (AI) to manage talent. Executive search and leadership consulting firms, as well as in-house talent acquisition teams, are interested in harnessing the data and automation to create a more effective, fair, and efficient hiring process. However, the use of artificial intelligence in human resources continues to raise problems ranging from data privacy and bias to legal and operational risks.
How do organizations overcome such challenges and harness the power of AI in their hiring processes?
Senior Fellow and Program Director of Human Capital at The Conference Board, Stela Lupushor, recently assisted in answering this question by contributing to the World Economic Forum's (WEF) toolkit on the responsible use of AI in the field. Lupushor will also be a featured panelist at AESC's global conference ADAPT: 2022 speaking alongside SHRM Chief Knowledge Officer Alexander Alonso, Ph.D., SHRM-SCP, and Accenture's Managing Director of HR Digital Transformation and People Analytics Colin Anderson, about the practical toolkit and the use of human-centered AI for human resources.
As an introduction to the topic and a sneak peek at what's to come at the conference, AESC spoke with Lupushor for an in-depth interview.
AI is increasingly being used in the workplace. How do you see AI transforming the talent landscape?
Let's start by discussing the impact alongside some definitions since many technology providers are claiming to use AI that might be either oversimplifying or overpromising the capabilities of their offerings. There are several ways to look at technology use in the workplace:
Automation that enables workers to perform activities faster or it completely takes over those tasks: This is typically a program (code) that contains step-by-step instructions and rules to guide decisions. Such programs are quite simplistic, rigid, lack contextual understanding and are unable to determine any secondary implications. For example, automated solutions review resumes and identify candidates that are a great match to the job opening based on the job description and matching keywords in that description to the candidate profile. Such solutions have been around, and their use will continue to be broad to automate routine activities, reduce error rates and simplify the workflow for the users.
Such solutions could transform or completely take over specific activities and as a result, fundamentally change the talent capabilities needed to perform the work. We are talking about a collaboration between people and machines which requires different attitudes, aptitudes and management skills to orchestrate the complexity. Routine or unsafe activities can be taken over by AI-enabled robotics; labor-intensive and time-consuming work can be streamlined and automated so talent can focus on complexities where machines could add little to no value (creative work, human interaction and support, teamwork etc.).
Augmentation of human decision-making where AI solutions enable significantly more complex actions: These can improve or change the process through predictions, create new data or augment existing ones with new kinds of information to support decision making. It can even introduce new tasks or perform existing ones in fundamentally different ways. It can also optimize for specific goals. For example, it can look at the applicants and predict their performance or likelihood of attrition or recommend edits to the job description that will make it less biased towards people earlier in their careers or certain characteristics that will more likely be represented in certain underprivileged groups.
Most AI-based HR tools, especially in the context of HR processes, use some form of machine learning, which is an approach that identifies patterns in training data including past examples of tasks and past outcomes. It assumes that these patterns will hold when applied to new cases in a system. For example, a machine learning algorithm designed to predict high potential job candidates might look at historical records of previous hires and look for patterns in the types of hires and their characteristics that correlate with better performance. In this case, the training data contains records of past hires and detailed personal and professional characteristics inferred from their resumes and job application, as well as their performance data (outcomes).
Let’s explore the impact through a specific segment: women 45-plus. If one considers the differences in lifecycle patterns of women, especially later in their career, someone who is over 45 years old—with multiple career gaps (to take care of children or aging parents) and multiple career pivots to gain necessary flexibility or support their partner's career—will rarely end up in the training data correlated with high performance. As such, AI is now being used to codify the biases we already have in society and it is a dangerous road, especially when it comes to employment decisions. Having women involved in the design process, deciding what training data is used and for what context, how it is processed and optimized, and interrogating the results for potential bias could bring the necessary counterbalance. Not to mention that involving women in such lucrative occupations will provide the so-needed financial security and job stability.
Executive search and leadership consulting is a profession built on strong, interactive, human-centered relationships. How would you suggest AI be used in the profession? And what would you consider as the drawbacks?
In the case of executive search and leadership consulting, AI can augment or automate time-consuming activities and leave human relationship management to humans. Examples that come to mind are finding the candidate that might closely match the job requirements. This involves parsing through a variety of information about candidates (including their resumes) to identify and score the match. It can also include a variety of assessments of skills and capabilities (through gamified experiences, self-recorded video interviews, or a self-exploration of interests, strengths and deficiencies) that will help determine the organizational fit and validate the candidate’s abilities at the pre-hire stage.
What should leaders consider when selecting the right AI solution to assist in finding and assessing candidates?
The way to select tools is to look at them through the risk management lens and as yourself questions such as:
- What is the intended use of the tool?
- Are there potential ways to misuse it?
- What will its impact be on organizational behaviors (Will humans try to “game” it)?
- What are the consequences for those who do not comply or perform as expected by the system?
Leaders should always assume that machine learning-based AI tools can introduce or perpetuate the bias (either through the data used to train the algorithm or through the developer’s worldview) and that whatever human behavior tends to be, the tool will learn and replicate it. It is therefore critical to involve in the tool selection as broad of a set of people to understand and evaluate different perspectives.
You'll be speaking at AESC’s May conference, Adapt: 2022. How will human-centered AI help search firms, CHROs, internal talent acquisition teams and business leaders adapt to today's environment?
We are at a unique stage of HR evolution and technology. AI solutions are creating an unprecedented degree of transparency.
Furthermore, many governance-monitoring agencies such as the Sustainability Accounting Standards Board (SASB), International Integrated Reporting Council (IIRC), and World Economic Forum (WEF) have called for qualitative and quantitative information about selected aspects of human capital performance as a proxy for organization’s governance quality. Combined with more recent human capital disclosure requirements from the SEC and the ISO 30-414:2018 standard – Human resource management, these will encourage companies to disclose and bring transparency to their workplace practices and outcomes.
Such transparency brings the need to address the issues that previously might have been kept close to the vest. The information about working conditions, culture, workplace practices, etc. permeates organizational boundaries and informs the prospects and employees alike. No longer can leaders think about solutions in an isolated fashion. It's now something for talent acquisition or leadership development teams to sort out. Solutions must be explored throughout the end-to-end employee experience.
Some of the opportunities for leaders to consider:
- Examine the candidates’ pipeline and understand if you are intentionally inviting and selecting candidates from underprivileged or under-tapped groups. Are you looking at different sourcing channels and understanding the performance of each, and considering the unintended consequences of channel choices (i.e. if you are advertising your jobs on Snapchat, your candidate pool will most likely be skewed towards younger, early in their career workers)? How many diverse candidates make it to interviews and how many get down-selected (this could point to the manager’s biases and inform you that a bias training might be needed)? One of the great advantages of technology in the past 16-18 months is that it allowed many to continue working remotely and proved to leadership that not every job needs to be done in person. This not only allows organizations to tap into talent that previously was not considered, but also brings more balance, especially for women, and gives them the ability to juggle work and family.
- When it comes to the development of talent, who are you investing in, and are those disproportionately individuals considered “high potential?" This will again more likely be skewed towards younger generations. All age and gender groups need to be developed and invested in. The data can shed light on the bias. The technology can also be a wonderful enabler to scale the development programs from resource-intensive and high-priced to a wide selection of online and asynchronous learning opportunities.
- Compensation practices and resulting pay disparity can be easily understood by looking at the compensation data between different groups and levels. Examining it over time will also shed the light on how your pay practices are applied and impact the income trajectory for different groups or segments of the workforce.
- Who are you retaining? Critically assess who is on the retention list and understand who you are predominantly optimizing for (which workforce segments end up being advanced or rewarded). Are you primarily retaining younger workers? Are there sufficient older workers, and especially older women to serve as role models or mentors to younger female employees? Like the hiring example, technology can enable organizations to retain women who need more flexibility in their schedules and allow them to work remotely or on a part-time basis without stepping out of the labor market altogether.
- How flexible are your policies, schedules and benefits? Are you inclusive in your design? For example, by adopting parental leave, you are enabling only those with children to take time off whereas adopting a caregiver leave will expand the number of workers who can take advantage of such leave.
- Lastly, how are you structuring the ways of working? Are you using technology to enable team collaboration and transparency in communications? Are you sufficiently investing in the skill-building necessary to use such technologies? Many of the workplace tools were designed by engineers for engineers and might not be as easy to adopt without adequate training.
When it comes to AI use specifically, it can influence the entire end-to-end employment journey as well. It can monitor, support automation and decision-making, and enable leaders to create a human-centric organization.
This is just a taste of what you can look forward to learning at AESC's global conference on May 11, 2022. Want to learn more about human-centered AI for human resources?
