COULD VARIED DATASETS AND ALGORITHMIC TRANSPARENCY INCREASE CONFIDENCE IN HR AI TOOLS? 

While the HR industry has been slow to invest in new technology — often looking for ROI metrics from first movers — the tight labor market and skills gap may prove a great motivator against risk-aversion. 

New research out of Wharton shines a light on the potential pitfalls of implementing AI to augment HR functions. The challenges that exist with traditional methods used to assess company and job fit can be exacerbated by tools that leverage backward-looking siloed datasets. Effectively, each database holds a clue that sheds light on the question of candidate fitness and allowing the data between these systems to “talk” to each other potentially increases reliability. In other words, the whole is greater than the sum of its parts.

https://knowledge.wharton.upenn.edu/article/using-ai-in-human-resources/

As more employers acquire AI-driven tools, underpinned by larger and varied datasets with the goal of a more effective selection process, the need for greater understanding into how those decisions are made — explainable and defensible AI — may increase focus on biases and unintended consequences. 

“To prevent algorithms from carrying harmful biases — like racial prejudice — decision-makers need to carefully assess how diverse their data collection is, and what values it reflects. Building an AI application starts with understanding the objectives, designing a hypothesis and identifying the necessary data to build the right algorithm. Then, in the final step in the process, there’s a great opportunity to identify bias.”

https://www.ciodive.com/news/how-ai-learns-the-biases-of-its-creators/563089/