AI in Talent Assessment: A Promising Mix, With Due Caution

Artificial intelligence (AI) technology has permeated a wide range of industries, with talent assessment being no exception. Broadly speaking, the practical goal of AI applications is to simulate human performance across various domains and tasks. Without a doubt, strong and increasingly generalized AI models have already profoundly changed the way we work and interact with technology, from enabling enhanced productivity through natural language informational queries to generating creative content across multiple forms of media.

3 Applications for AI in Talent Assessment

Below, we outline three promising applications for AI in talent assessment, followed by a summary of considerations that might necessitate prudent caution when using AI in talent assessments. Please note that these are examples, not an exhaustive list or discussion.

1. Improving Efficiency

A common use of AI thus far has been to assist in the generation of client-facing content. This includes generating content ideas as well as drafts of the content itself, such as lists, blogs, emails, and other resources or communications. In the context of talent assessment, AI can help generate initial item pools for psychometric instruments or create summaries of large volumes of text. With its natural language processing capabilities, AI can also help with generating insights from large qualitative datasets, which might otherwise require many personnel hours to parse and summarize. Finally, AI can enable more “natural” interactions with quantitative outputs of talent assessments, turning what might otherwise be a series of clicks or lines of code into a simple conversational query.

2. Advancing Data Collection and Predictive Power

AI may also be able to support talent assessment by providing a variety of new predictive tools that traditional statistical models cannot offer. For example, in asynchronous interviewing, AI could facilitate faster and more robust scoring of interview responses with greater consistency than a human counterpart. It should be noted that while this application of AI is possible in principle, it depends upon a high quality validated underlying model. Lastly, if a model is programmed to learn and update autonomously, scoring algorithms and databases of validity data could be updated much faster and more regularly than traditional methods allow.

3. Improving Customer Experience

Customers purchase assessments and pay for talent development services with the goal of obtaining actionable data and insights. AI technology has the potential to both speed up the processing of assessment data and provide increased granularity in generated output. As a result, the interpretive feedback associated with many talent assessment and development products could become more customized, intuitive, and informative on a case-by-case basis, without requiring human intervention, such as analysis and reporting. Additionally, because language processing is central to many AI models, customers could receive automated answers to conversational queries about their assessment results.

Caution #1: The Risk of False Information

When implementing AI as part of a workflow, it is imperative that all outputs be verified by subject matter experts (SMEs). While it can be easy to get caught up in the speed and intuitiveness of interacting with an AI model, we must remember that these models can make mistakes, and it is our responsibility to catch them. This is particularly important when using AI to augment research and development efforts, where correct answers may not be immediately apparent and cross-validation data is unavailable. There have been many instances of generative AI tools presenting information as fact that either does not exist or is categorically false.

Caution #2: The “Black Box” Problem

AI’s exciting progress shouldn’t overshadow the need for measured, thoughtful steps forward. Evidence-based practitioners who work in the development and application of assessments rely on a wealth of academic literature to inform psychometric and management consulting decisions. Evidence for the validity of interventions and assessments is, for the most part, rooted in traditional statistical methods. This allows practitioners to understand the inner workings of assessments and trace any generated output back to individual item responses. However, in the case of AI models, absent highly qualified technical staff, assessment providers may run the risk of not fully understanding how these tools actually work. This has serious implications, including reducing the ability of SMEs to validate tools and potentially harming the legal defensibility of offerings. Therefore, it is imperative that any integration of AI be carefully documented and that quantitative models are appropriately restricted to avoid turning products into a proverbial “black box” — a system, process, or model whose internal workings are not visible, understood, or accessible to the user.

The Big Picture: AI in Talent Assessment

AI technology is advancing rapidly and is being adopted for a wide variety of use cases. While AI has the potential to greatly improve the quality of talent assessment offerings and the speed with which they are generated, it is important to remember that its application in this space remains nascent. By balancing technical innovation with research and due caution, we can maintain progress in this area while continuing to deliver value to our clients.

Looking for More?

SIGMA offers a range of assessments that can help you quantify leadership, cognitive ability, personality, emotional intelligence and more. Our tools are scientifically validated, and industry proven. We have conducted more than 4 million assessments and received nearly 4,500 academic citations. If you are interested in learning more about our assessment solutions, complete the form below to tell us more about your needs and we will contact you shortly.