Artificial Intelligence in the Hiring Process

Table of contents

• Introduction
• Use of artificial intelligence by hiring managers in the hiring process
• Use of artificial intelligence by candidates during the hiring process
• Key takeaways
• Other resources

Introduction

Using artificial intelligence (AI) during the hiring process can open up many possibilities for hiring managers and candidates, but it also raises some important considerations.

This guide is intended to equip hiring managers to use AI in the hiring process within the framework established by their deputy head, who remains accountable for all hiring within their organization.

It provides context on how key requirements apply when AI is used in the hiring process. These requirements are found in the Public Service Commission of Canada (PSC)’s Appointment Framework and in other instruments, such as the Treasury Board’s Directive on Automated Decision-Making and the Treasury Board of Canada Secretariat’s Guide on the use of generative artificial intelligence.

This guide is evergreen and will be updated to respond to changes in the rapidly evolving AI landscape, and as new tools are developed and used in the hiring process.

Use of artificial intelligence by hiring managers in the hiring process

As a hiring manager, you’re accountable for all decisions made during an appointment process. You’re also responsible for ensuring that subdelegated appointment and appointment-related authorities are respected.

If you choose to use AI in the hiring process, remember that you must:

• understand the AI systems you plan to use
• be able to explain the decisions you’ve made throughout the appointment process
• especially for informal discussions or for recourse purposes
• validate AI outputs, including AI-generated ideas and suggestions to ensure that the AI-generated content is accurate and relevant
• this includes reviewing and adapting AI input to more accurately reflect hiring needs, and to remove biases and barriers or mitigate their impact
• ensure that the information conveyed in English and in French is the same throughout the hiring process because AI tools can produce translation errors
• this will ensure that candidates are not disadvantaged based on their first official language or the official language in which they choose to be assessed during the hiring process
• be transparent and open: let candidates know when and how AI will be used during the hiring process; this includes informing them when they’re interacting with a chatbot or a virtual assistant
• comply with the security, confidentiality, data protection and intellectual property requirements related to AI tools

Important

Whenever you use AI to recommend or support any staffing decisions, you must comply with the Directive on Automated Decision-Making’s requirements. This means you’ll need to:

• conduct an Algorithmic Impact Assessment before using any automated decision system
• provide notice to candidates that the decision will be made or assisted by an automated decision system
• provide candidates with a meaningful explanation of how and why the decision was made
• tell candidates how to proceed if they want to raise concerns about the decision or challenge it; recourse options must be timely, effective and easy to access

Establishing qualifications

Green checkmark You can use AI to help establish the qualifications required for a position. You’re responsible for the qualifications, and you must ensure they align with the position’s requirements.

Considerations

• AI tools can help you:
• think through qualifications for your position, thereby reducing time during this step of the process
• write and edit the qualifications in plain language
• AI tools can also:
• misinterpret specific job requirements or fail to consider the organizational context
• introduce or create biases and barriers that could disadvantage people belonging to equity-seeking groups (for example, by requiring consecutive years of experience or recent experience)

Job advertisements and communication with candidates

Green checkmark You can use AI to develop job advertisements and communication products for candidates (invitations to tests and interviews) and to interact with candidates in an automated way (chatbots or virtual assistants).

Considerations

• Using AI to create invitations and messages can help you manage communications with candidates efficiently while tailoring messages to the recipient, creating a more personalized, engaging and inclusive experience
• Whichever AI tool you use, you’re responsible for the accuracy and quality of communications
  • You must tell candidates when they are interacting with a chatbot or a virtual assistant
• AI can overlook contextual and cultural nuances
  • Consider using a collaborative review process that involves hiring managers and diversity and inclusion specialists
    • This helps you adapt AI-generated content and ensure that communications are appropriate and attuned to cultural and contextual nuances

It’s strongly recommended that you indicate in the job advertisement whether candidates can use AI when applying for the job (for example, to generate answers to screening questions, to write a cover letter). This is particularly important if written communication will be assessed during the screening stage.

Developing assessment materials

Green checkmark You can use AI to develop assessment materials (for example, interview questions or scenarios).

Important

Before using any assessment method, you must identify any biases and barriers^{Footnote 1} and implement strategies to remove them or mitigate their impact.

Considerations

• Use of AI may reduce the time required to develop assessment material
• You’re responsible for using tools that provide for fair assessment and accurately measure the position’s qualifications
• Candidates could use AI tools to predict test or interview questions and prepare answers
  • Check whether AI can be easily used to obtain the expected answers; if so, adjust the assessment strategy to ensure you’re evaluating the candidate’s qualifications rather than the AI-generated response

Assessing candidates

Important

Before using any assessment method, you must identify any biases and barriers and implement strategies to remove them or mitigate their impact.

Before the assessment(s), you must let candidates know which methods or tools will be used and advise them of the opportunity and process to request accommodation measures.

Green checkmark You can use AI to administer assessments (for example, platforms allowing for remote supervision, asynchronous interview tools).

You can use AI tools to assess candidates (for example, automated ranking or sorting of candidates during screening, or automated scoring of tests or interviews) only if you can clearly explain the following: 

1. The role the AI tool played in your decision process
2. The criteria, data or mechanisms that the AI tool used to assess the candidates’ answers
3. The assessment performed or individual feedback provided by the AI tool for each candidate
4. The way you interpreted and used the AI system’s output when making decisions about candidates.

Considerations

• Using AI tools to assess candidates poses a risk because these tools can be complex and difficult to explain
• Regardless of the AI tool used, you’re accountable for the decisions made during the hiring process and for the final decision about candidates
• You must document the required information in accordance with Annex B of the Appointment Policy

Icon of a red ”x” Inappropriate use of AI to assess candidates

Example 1: An AI tool is used to rank and sort applications based on the information contained in resumés.

The tool uses non-transparent criteria, such as specific keywords or biased historical data models.

Each candidate receives a score based on unknown parameters, making it difficult to understand the scoring.

Recruiters interpret the scores without understanding the underlying mechanisms, which can lead to biased or unfair decisions.

Example 2: An AI tool is used to conduct video interviews with candidates. AI analyzes candidates’ verbal answers, non-verbal communication and appearance to assign scores and provide recommendations to hiring managers.

Candidates or hiring managers don’t fully understand the impact of AI on final decisions. The assessment criteria, the data used (for example, facial expressions or tone of voice) and AI mechanisms are not transparent.

Candidates receive AI-generated scores or comments, but these assessments are not clearly explained. For example, a candidate might be told that they have a low communication score without being told which specific aspects of their performance impacted their score.

When making decisions, hiring managers will take AI recommendations into account, but they won’t always understand how these recommendations were produced. This can lead to biased or misinformed decisions.

Use of artificial intelligence by candidates during the hiring process

Green checkmark You can allow candidates to use AI throughout the hiring process.

The use of AI could be relevant if access to these tools is part of the job. Allowing candidates to use AI tools during the hiring process can:

• simulate a realistic work environment
• better reflect the professional context in which these tools are integrated
• allow you to assess their ability to use them effectively

Whether you choose to allow or prohibit the use of AI by candidates, you should:  

1. Develop the assessment strategy, methods and tools, remembering that candidates have easy access to AI. This can involve:
   • using other assessment methods, such as an interview, reference questions or scenarios to confirm that the assessment reflects the candidate’s qualifications rather than AI-generated information 
   • asking the candidates to use generative AI to assess skills related to using such tools (for example, checking the accuracy of results, developing appropriate prompts)
2. Give candidates clear rules and communicate the possible consequences of breaking these rules (for example, invalidation of test results, elimination from the hiring process, possibility of an investigation)
3. Ask candidates to confirm in writing that they have read and understood the rules and consequences.

Allowing candidates to use AI 

• Specify which types of resources and tools candidates are allowed to use and the stage(s) of the hiring process during which they can use them (for example, when answering screening questions, writing cover letters, completing unsupervised tests)
  • Various factors can create unequal access to AI tools during the hiring process: access to paid versions, the language of the tool, candidates’ technical knowledge and the available computer equipment
• Clearly state that candidates must cite any AI tool used by identifying the AI-generated content and its source
• Explain the requirements for protected material (specify that candidates are not allowed to share secure test material or to submit content to unsecured AI tools)

Sample statement: The use of AI is permitted

The use of Artificial Intelligence (AI) tools (for example, Copilot, ChatGPT or any other form of AI) is permitted. However, you must explicitly disclose when you use such tools in your answers by identifying the AI-generated content and its source. You may be asked to explain or elaborate on your answers in a follow-up interview. If you fail to disclose your use of AI tools or if you use AI tools inappropriately, it may result in an investigation and serious consequences, including the rejection of your application.

*By submitting this assessment, I acknowledge that I have read, understood and accept the integrity terms outlined above. I certify that I have clearly disclosed any use of AI and identified the AI-generated content and its source. I understand the potential consequences that failing to meet these requirements may have on my application.

Prohibiting candidates from using AI

• When assessments are administered in an unsupervised environment, the risk of a candidate using unauthorized resources, including generative AI tools, increases significantly
  • Consider remote supervision (monitoring assessments online through video, audio, and/or screen sharing to ensure integrity) in order to encourage candidates to use their own skills and to increase fairness
  • Consider using a range of assessment methods (such as structured interviews and reference checks) and approaches (for example, unsupervised followed by supervised) to provide a more comprehensive view of candidates’ qualifications, reduce reliance on AI, and minimize potential biases and barriers.

Sample statement: The use of AI is prohibited

You must complete the assessment independently without external help, including unauthorized internet resources or Artificial Intelligence (AI) tools (for example, Copilot, ChatGPT or any other form of AI). You may be asked to explain or elaborate on your answers in a follow-up interview. Any violation may be investigated and result in serious consequences, including the rejection of your application.

*By submitting this assessment, I acknowledge that I have read, understood and accept the confidentiality and integrity terms outlined above. I certify that I have completed this assessment independently without external assistance, and that I have neither plagiarized nor used AI. I understand the potential consequences to my application if I fail to comply with these requirements.

Unauthorized use of AI by candidates

Using AI detection tools to determine whether candidates have used AI

• If you use AI detection tools to determine whether a candidate used AI, it’s considered an automated assessment of candidates:
• the requirements of the Directive on Automated Decision-Making apply
• you must explain to candidates the factors that led you to use the tool
• Be careful if you choose to use AI detection tools, as their reliability may be questionable
• A false positive could lead to serious consequences for candidates; it may be more prudent to reassess the candidate

Other ways to detect the unauthorized use of AI by candidates

Be on the lookout for the following signs (not proof) that AI may have been used:

• some answers are written at a level or in a style that is significantly different from other answers
• answers are generic, don’t directly address the subject or don’t make sense for the scenario
• answers include fake references or studies
• answers describing the candidate’s work experience are inconsistent with their employment history
• an answer is written in the third person (when the candidate was asked to describe their experience)
• in pre-recorded videos, the candidate appears to be reading from a script
• answers follow the typical response pattern of AI software

What to do if unauthorized use of AI is detected

Concerns that a candidate has cheated or failed to comply with assessment instructions may be investigated as improper conduct.

If unauthorized use is detected during the process (before any appointment), the irregularity can be addressed informally without an investigation (for example, by eliminating or reassessing the candidate on a case-by-case basis).

The PSC may choose to examine certain instances of improper conduct as fraud. If this is the case, you can still address the irregularity informally as improper conduct (there’s no need to wait for the outcome of the investigation of fraud).

When to refer to the PSC for an investigation of fraud

Situations can be referred to the PSC for an investigation of fraud when the instructions explicitly prohibit the use of AI and you have reason to believe that AI was used.

Key takeaways

• AI has the potential to make the hiring process more effective and efficient
• AI tools can be helpful resources, but they should be used with caution and diligence
• Their features and capabilities vary
• The choice of tool will affect the results’ validity and accuracy
• If  you use a generative AI tool, following the principles and best practices in the Guide on the use of generative artificial intelligence is strongly encouraged
• Using AI tools to assess candidates poses a risk because these tools may be opaque and complex; only use them if you can clearly explain your decisions
• If you use AI tools to assess candidates, you must comply with the requirements of the Directive on Automated Decision-Making. This means you’ll need to:
  • conduct an Algorithmic Impact Assessment before using any automated decision system
• provide notice to candidates that the decision will be made or assisted by an automated decision system
• provide candidates with a meaningful explanation of how and why the decision was made
• tell candidates how to proceed if they want to raise concerns about the decision or challenge it; recourse options must be timely, effective and easy to access
• Using AI detection tools to determine whether a candidate used AI is considered an automated assessment of the candidates, and the requirements of the Directive on Automated Decision-Making apply
• Before using any assessment method, you must identify any biases and barriers and implement strategies to remove them or mitigate their impact
• Whether or not you choose to allow candidates to use AI, be sure to give them clear rules and communicate the possible consequences of breaking these rules
• For example, invalidation of test results, elimination from the hiring process, possibility of an investigation
• You can also ask candidates to confirm in writing that they have read and understood the rules and consequences

Other resources

Support

Hiring managers can contact their human resources department to discuss how certain issues are handled within their organization.

The PSC provides support through staffing support advisors designated for each department and agency.

Training

The Canada School of Public Service offers a variety of courses to help you learn about AI. Explore the school’s learning catalogue for courses, events, programs and other learning tools, as well as more information on their products and services.

Useful links

• Directive on Automated Decision-Making 
• Guide on the Scope of the Directive on Automated Decision-Making 
• Generative AI in your daily work 
• Responsible use of artificial intelligence in government 
• Guide on the use of generative artificial intelligence 
• The Future of Generative AI
• Best practices for unsupervised testing
• Public service hiring guides
• Guide to Mitigating Biases and Barriers in Assessment
• Directive on the Management of Communications and Federal Identity