Navigating the hype and impact of AI in HR

In a recent report (Artificial intelligence in human resources management: Challenges and a path forward) on the impact of artificial intelligence in HR by Peter Cappelli, Prasanna Tambe, and Valery Yakubovich, the authors examine the challenges, current innovation process, and some of the vendors that are using AI to disrupt HR practices.

The report examines four challenges in using data science techniques in HR practices: 1) complexity of HR phenomena, 2) constraints imposed by small data sets, 3) ethical questions associated with fairness and legal constraints, and 4) employee reaction to management via data-based algorithms. The report outlines a conventional AI Life Cycle as follows:

The life cycle of an AI-supported HR practice

The life cycle of an AI-supported HR practice

Operations: With approximately 60% of all business spending in the US is on labour costs, it’s no wonder that AI is being used to complete transactional requirements and process improvement. Operations areas such as recruitment and performance management produce large volumes of data that lead into the next stage - data generation.

 “One of the reasons for the interest in applying data science tools to human resources is because HR performs so many operations and so much money is involved in them”

Data generation: Though HR has considerably less data then a function like marketing, there are still many HR activities that create a lot of data exhaust – or digital information – that can be used to help develop algorithms. Typically these inputs have to be joined together or extracted from multiple databases before any analysis can take place.

Machine learning: This is a set of techniques that can adapt and learn from data to create algorithms that perform better and better at a task over time. In HR the most common application of machine learning is in “supervised” applications, in which a data scientist creates a machine learning algorithm, determines the most appropriate metric to assess its accuracy, and trains the algorithm using the training sample.

Decision-making: This is the final stage where the insights gained from the above processes are used to support and influence daily decisions.

Recommendations for using AI in HR

Each stage has its own particular set of challenges that need to be addressed. For example, the ethical collection of data, explainability of predictions, managing employee expectations, and gaining buy-in and trust. The report outlines various challenges at each stage of the life cycle and makes a number of recommendations, including the following when focusing on the data generation stage :

  1. Don’t expect perfect measure for things like performance as they don’t exist. It is better to choose reasonable measures and stick with them to see patterns and changes in results than to keep tinkering with systems to find the perfect measure.

  2. Aggregate data from multiple sources and perspectives and do this over time. The benefit of so many of the newer digital HR tools that are now available is how easy it has become to build new datasets and gather user input. However, this also adds complexity in how to integrate all of the new datasets, making database management a critical component for HR moving forward.

  3. Integrate HR data with company data, such as finance, to monitor the effects of HR practices and the impact on business outcomes.

With the amount of available organisational information, it can be difficult for HR managers to determine what data needs to be included in early stages of analysis. This is where getting feedback from the business on the critical issues that they are facing can be used to develop a hypothesis to be tested using the data.

Other interesting points from the report include:

Small data: Most organisations are not large enough to collect enough data to develop a predictive algorithm. However, this doesn’t mean that the data collected is useless, rather it can be used to determine causal relationships – for example, the number of qualified applicants from specific sources.

Employee reactions: There is an increasing use of employers using internal and external data to identify candidates, bad behaviour, and flight risks. Emails scanning for harassment and bias behaviour. This monitoring is not new. However the extent that HR can monitor employees is increasing such as being able to determine sentiment of messages posted on social media or emails sent. The boundary of what is ethical and what is not is blurring and bending.

Fairness: There are reasonable concerns about using algorithms for decision-making, especially given that most are backwards looking. An example of this is Amazon’s hiring algorithm that was inadvertently biased towards women. Though algorithms have the opportunity to improve on hiring managers’ biases, the balance between appropriateness and predictive power needs to be weighed to determine potential adverse impact. Algorithms also need to provide explainability to enable the user to understand the decisions made.

Conclusion

There is endless potential for using AI to augment HR practices, though in the process of adoption, considerations need to be made. The challenges outlined in the report are not meant to block or hinder this process, but rather provide the necessary argument for the safe and ethical deployment of AI in HR.

This report comes at an interesting time, as many discussions about the application of AI in HR turn towards ethical and reliability concerns. This is another indicator that AI technology is HR is jumping the chasm to mainstream adoption. Though these discussions are essential and required, it can cause some to be put off by the technology.

There is currently a void in laws regarding the use of AI in business and HR – this is where the community of vendors, HR professionals, academics, and thought leaders can take the lead in developing industry best practices and hosting these crucial discussions. However, this is not an indicator to not explore and experiment with AI in HR, but rather an indicator to ensure that your organisation is having the right discussions, building development roadmaps, testing products, and thinking of creative and new applications of AI. It is also important that while doing this, organisations talk about ethics early in the process and don’t try to tack it on later.

If you want to stay up-to-date on the latest trends and news in AI and the Future of Work, then you can subscribe to the CognitionX weekly newsletter on this topic.


ABOUT THE AUTHOR

Megan Marie Butler is a research analyst at CognitionX, a leading AI research house and advice platform. Megan specialises in research on the impact of AI on HR and through her professional and educational work (she is studying for a PhD), Megan has developed a passion for HR technology and the impact that both people and the technology they use can have on a company. You can find other articles like this on Megan’s medium page.