Creepy Analytics: Avoid Crossing the Line and Establish Ethical HR Analytics for Smarter Workforce Decisions by Salvatore V. Falletta.
Salvatore holds a EdD in Human Resource Development from North Carolina State University and a Master of Public Administration with a specialty in HR and Organization Development from Indiana State University. He previously served as Vice President and Chief Human Resources Officer for Atmel Corporation and held leadership HR roles at Intel, SAP, and Sun Microsystems. Today Salvatore is a Professor of the Practice at Vanderbilt University.
Salvatore is revealing while AI can certainly augment an HR Recruiter’s efficiency, it simply cannot operate in a vacuum. Similar to many authors, a Human In The Loop (HITL) is absolutely mandatory to prevent a datafication of people from turning into surveillance.
Similarly, his golden rule is that AI should never make a final workforce decision alone. This is understood as ‘Augmentation, Not Replacement’ AI should be used to scan thousands of resumes, but human recruiters must provide the final judgment.
Its all about the power of Regex
Salvatore sees that organizations must move away from “black box” algorithms where the decision-making process is hidden. Through his Seven-Step HR Analytics Cycle, readers will understand “black box” algorithms to transparent, evidence-based practices that prioritize data relevance over vast amounts of data collection.
This is an interesting take when written in February 2024 as the role of Agentic AI was already emerging. Many early use cases are in HR and even in the key elements described below:
- Explicit Disclosure: Candidates should be informed when AI is being used to evaluate them.
- Explainability: Recruiters must be able to explain why an AI tool recommended a specific candidate in plain, non-technical language. If you can’t explain the logic, the tool is likely “creepy.”
- Real-time Voice: Use AI to give candidates a “voice” (e.g., through feedback loops) rather than just treating them as data points to be filtered.
Perhaps Agentic AI’s orchestration changes his position today for the better. Agents are simply executing tasks assigned by HR in order to streamline efforts of understaffed HR teams.
He does outline needed ethics to begin training the AI service, not just the output. He is addressing how organizations must “Understand the DNA” which includes historical data. Naturally if the training data is based on past biased hiring practices, the AI service will simply automate that discrimination. The good news is that he writes about the Relevance Over Availability, Facial expression analysis and social media scraping which all negatively impact the HR outcomes
In conclusion, Creepy Analytics provides readers a roadmap for navigating the ethical minefield human resources. He warns against the “creepy” surveillance that destroys employee trust, advocating instead for a “human-in-the-loop” approach.