Government Publishes Guidance on Responsible AI in Recruitment Processes

The Department for Science, Innovation & Technology (DSIT) have published guidance relating to “Responsible AI in Recruitment” specifically relating to the use of artificial intelligence (AI) in the use of AI tools in HR and recruitment processes including screening, selection, interviewing and other automated processes. 

The guidance is aimed at removing unfair bias or discrimination that may exist in the tools used for recruitment and selection by HR departments, as well as removing barriers from digital exclusion where applicants may not be able to use technology for a variety of reasons such as access or proficiency. It provides guidance on procuring AI systems from third parties and ensuring organisations consider the risk posed by the systems as well as identifying appropriate assurance mechanisms during and after procurement.  

What does the guidance say?

The aim of the guidance is to set out steps for organisations to consider when procuring, deploying and using AI in recruitment to ensure the UK Government’s AI regulatory principles, outlined in ‘A pro-innovation approach to AI regulation’ are followed, being:

• Safety, security and robustness;
• Appropriate transparency and explainability;
• Fairness;
• Accountability and governance; and
• Contestability and redress,
• together, the Principles.

The guidance provides steps and potential questions to ask before and during the procurement process. Questions for consideration at each stage are provided with examples of good practice and key points to take into account when devising quality questions. 

The guidance also sets out assurance mechanisms for each stage which the organisation may choose to use all or some of. These are used to address the considerations, aide the decision-making process and ensure compliance with the Principles. 

The relevant considerations and assurance mechanisms for each step are set out in the table below:

Before Procurement
Considerations	Assurance Mechanisms
what is the desired purpose?  what are the desired outputs?  how will the system integrate with existing software? are reasonable adjustments required?	algorithmic impact assessment equality impact assessment  data protection impact assessment AI governance framework
During Procurement
accuracy and scientific validity risk identification and communication	bias audit  performance testing risk assessment model cards
Before deployment
avoiding incorrect usage assess performance against equalities outcomes plan reasonable adjustments	performance testing  training impact assessments  transparency
Go live
ongoing monitoring contestability and redress	iterative performance testing  bias audits user feedback

 

Our views

The guidance is seen as a framework to enable procuring organisations to ensure relevant requirements and procedures are in place to ensure accountability, review and compliance with the Principles when procuring AI tools in an emerging technological market. Whilst the procedures and processes set out in the guidance are aimed at HR and recruitment, they can be used for any AI procurement and may already be used in practice to some degree. 

Each consideration and assurance mechanism comes with examples and guidance to enable organisations to confidently assess software and tools and should be a good aide in procuring AI, designed with the organisation in mind.