What Organizations Should Include in Their Artificial Intelligence Policy [Part 2]

Estimated reading time: 8 minutes

While artificial intelligence (AI) may be very much within the news, the technology continues to be recent. It could actually be difficult to draft a policy about something that we’re still learning. 

In the primary article of this series on artificial intelligence policies, we discussed the explanations that organizations need to contemplate drafting a policy. In this text, let’s take a better take a look at what organizations should consider on the subject of creating policy. 

To assist us learn more, I’ve been chatting with our friend Carrie Cherveny, chief compliance officer and senior vice chairman of strategic solutions at HUB International. In her role at HUB International, Carrie works with clients to develop strategies that ensure compliance and risk mitigation on the subject of advantages and employment practices. 

Because we’re talking about human resources policy, please do not forget that Carrie’s comments mustn’t be construed as legal advice or as pertaining to any specific factual situations. If you may have detailed questions, they must be addressed directly along with your friendly neighborhood labor attorney.

Carrie, thanks again for helping us understand this topic. In terms of policy development, organizations have a chance to state their positions about something. For instance, organizations might discuss their commitment to moral conduct and compliance when introducing their code of conduct. Are there some things that organizations might want to contemplate confirming their position on when introducing an AI policy? 

[Cherveny] AI is revolutionary. With such dramatic change comes fear, uncertainty, and doubt. Compounding the concerns about AI is the shortage of transparency and visibility into the AI programming. There’s really no strategy to ‘look under the hood’ and inspect the AI engine. Consequently, there’s no strategy to know if the system was developed with any inherent bias. Furthermore, because AI is machine learning (meaning it learns from the top user), there’s no strategy to know if the AI is adopting an unconscious bias of the end-user. 

For instance, let’s say a recruiter is using AI to sort through candidate resumes and first interviews. The recruiter that selects candidates to go forward in the method has an unconscious bias and leans into younger females for the role. Is it possible that the AI will learn from the recruiter and likewise highlight younger females as ‘top candidates’ for the role?

To manage for these possibilities, employers must remember to at all times “pressure test’ the AI results. Within the recruiting example, the recruiter should frequently review the resumes and the candidate profiles for the rejected candidates or those on the underside of the list. 

Be your personal end-user. In other words, use the AI and be the candidate. Change your name to names which will reflect various ethnicities. Change your resume to reflect various years of service (i.e. age). Change your address to reflect various geographic locations. Do you get the identical result from the applying tracking system? 

The risks of a discriminatory result or process using AI also creates fear, uncertainty, and doubt on your employees. Employers have made great strides in the sector of diversity, equity, and inclusion – if not managed properly AI could create a major setback.  

The AI policy provides the chance for the employer to reassure employees that:

• A human will at all times be the decisionmaker
• The employer is committed to an environment free from discrimination
• The employer is committed to its diversity, equity, and inclusion (DEI) initiatives and programs
• The organization has robust AI audit processes to stop a discriminatory or result

For instance, since we’re talking about AI, I asked Microsoft CoPilot for ‘sample language for an AI policy reassuring employees that AI won’t ever make a final decision and HR will provide oversight to avoid discrimination or a disparate impact’. Here’s what was provided:

While we will’t tell organizations use AI – that can be driven by industry, jobs, etc. – it seems to me that any AI policy needs to present employees a strategy to ask questions or express concerns. Do organizations need to incorporate some type of reporting mechanism of their policy? And in that case, should this be reported to HR or possibly some type of “AI Committee”?

[Cherveny] Have you ever ever had a pleasing customer support chat-bot experience? After I ask conference attendees this query, I often receive a convincing, unanimous ‘No!’ or ‘Never!’. It’s one thing to be a frustrated customer, it’s one other to be an worker being denied their rights under various federal laws. 

An worker difficult chat-bot experience generally is a violation of varied federal laws. For instance, some AI tools may require verbal or video interactions. There are AI chatbots that may conduct a candidate interview or to help an existing worker with advantages or handbook questions. Likewise, employers may use AI video tools to conduct a candidate interview or conduct recent hire orientation. Using these tools is just not illegal and might often create significant efficiencies.

But – what in case your candidate or worker has an impairment that makes it difficult for the person to speak with the AI? For instance, a video AI tool may not provide a positive rating for a candidate with a speech impediment, strong accent, or a facial tick. Likewise, an AI chatbot may not provide a high rating for a candidate who has dyslexia. How can that candidate or worker get past your AI tool and reach a live person? 

These are only just a few of the examples that make it essential for employers to create an ‘easy button’ for candidates and employees to acquire access to a live person. There are at the very least two regulations which may be applicable here. 

The Americans with Disability Act (ADA): The employer relies on an algorithmic decision-making tool that intentionally or unintentionally ‘screens out’ a person with a disability, despite the fact that that individual is capable of do the job with an affordable accommodation. ‘Screenout’ occurs when a disability prevents a job applicant or worker from meeting – or lowers their performance on – a range criterion, and the applicant or worker loses a job opportunity because of this. A disability could have this effect by, for instance, reducing the accuracy of the assessment, creating special circumstances which have not been taken under consideration, or stopping the person from participating within the assessment altogether.

The Americans with Disabilities Act would require that the employer provide the candidate or worker with a disabling condition easy accessibility to an avenue to request an accommodation. For instance, a candidate with a speech impediment will need a simple strategy to request the corporate to offer a live human for the interview as an alternative of the AI. 

 Title VII of the Civil Rights Act: As in previous example, Title VII may apply if the candidate or worker doesn’t speak English as a primary language and/or could have an accent. Failure to present candidates and employees the identical opportunities no matter their national origin (i.e. their accent) may run afoul of Title VII if English proficiency Will not be a legitimate position requirement.  The employer must be sure that candidates and/or worker don’t suffer a disparate impact on the premise of national origin. 

The Equal Employment Opportunity Commission (EEOC) has been ahead of the AI curve and has provided useful and knowledge guidance on these topics.

My because of Carrie for sharing her knowledge with us. Organizations have lots to contemplate when drafting a synthetic intelligence policy. There are the considerations based in your industry and jobs. We talked about a few of those facets in the primary article. After which there’s existing laws, which is changing to satisfy the needs of the fashionable workplace. 

Along with the guidance being provided by the EEOC, remember to take a look at the checklist created by HUB International on how HR departments can seamlessly integrate AI into their workflows. And that’s going to steer us to our third (and final) article on this series on artificial intelligence – how can human resources departments effectively implement an AI policy of their company. 

Image created by DALL-E for Sharlyn Lauby

The post What Organizations Should Include in Their Artificial Intelligence Policy [Part 2] appeared first on hr bartender.