Why Businesses Should Develop an AI Ethics Policy
Instantprint, the UK's largest online printer, instituted an AI ethics policy. Its head of marketing says every business should.
More and more, companies are talking about the need for AI ethics policies or codes of conduct to provide guidance when faced with ethical decisions regarding the use of artificial intelligence.
Policies typically include principles such as fairness, transparency, privacy, accountability, and security, to align AI usage with the organization's values. Government agencies and international organizations have also released guiding principles and codes of conduct for AI to promote safe, secure, and trustworthy AI use worldwide.
Dan Robinson, Head of Marketing and eCommerce at instantprint, the UK’s largest online printer, recommends that businesses implement and enforce such policies to aid smooth AI implementation.
Remarking on how and why instantprint implemented an AI ethics policy (or code of conduct as he refers to it), Robinson said:
“Instantprint employees must use the platforms responsibly and adhere to the guidelines we’ve set out. We nurture an environment of trust but also provide our team with the rules and regulations they need to know to use these tools effectively and safely. Our “AI Code of Conduct” is set out by each platform we use, with dos and don’ts for each tool.
“We want to make our AI policy a collaborative space. Those using AI will get to know the platforms in great detail. Making policies a shared effort means that we’re more likely to have rules that will work for our team, developed by our team, with the exception of legal and ethical frameworks as a standard.
“Transparency is fundamental in our approach. Those using the tools daily have the biggest insight, openly sharing this information facilitates a better understanding and gives accountability, it also means that we can combat concerns faster.
“AI is wonderful, but like anything in the workplace, a policy is essential.”
The dos and don’ts Robinson refers to are as follows:
TRANSPARENTLY use AI and do not claim things as original or unique.
FAIRLY use AI in any decision-making process and remain unbiased.
QUESTION AND CHECK facts, claims, source origin, and intellectual property, use AI and edit.
NEVER share personal or identifiable customer or business information with AI tools.
REPORT anything that may seem suspicious or not quite right.
“For AI, we have strict rules in place about fact-checking. As a reputable business, we need to make sure all information we put out is true and from a reliable source. ChatGPT’s knowledge is limited past 2021, so we make sure our team is aware of this, and we always get them to check over the copy written by ChatGPT before publishing.”
Instantprint began its policy formation by cataloging and evaluating all the AI tools it uses. Following this, the company carried out a survey among its employees to gauge their views on AI and its impact on their work. The survey, which included over 1,000 office staff, aimed to identify concerns, spending trends, and examples of AI integration in the workplace.
The company maintains a working document, regularly updated with dates, to develop an adaptive policy that keeps everyone informed about the latest developments. Robinson notes that the process of drafting AI policies is ongoing, as learning about AI continues.
Dan Robinson Interview
Robinson was kind enough to allow us to interview him to learn more about his views and the reasons why he feels businesses should implement an AI policy.
Q: What areas (departments, divisions) should such policies cover, and in what respect? For example, HR would have different needs and goals than marketing or customer service and, therefore, needs a different policy scope.
DR: Policies should cover all departments but should be adapted accordingly. We predominantly use AI in our marketing team, but you can already see how it is growing across other areas of business, from AI chatbots in customer service to the use of AI to sort complex data.
I think having an overarching company policy is a great place to start. It should then fall to the responsibility of department heads to craft these into policies that suit the needs of their team and use case.
Q: Can you supply me with Instantprint's AI code of conduct? Would you suggest it serves as a model or template for others to use?
Our guide follows this structure:
AI Aims – the why, why do we use AI, what do we want to get from it and its commercial input.
Specific tools we use are then broken down into why, who, what, where, and when.
For example, ChatGPT: why do we use it, who can it be used by and use case examples, what can it be used for and use case examples including dos and don’ts, where we can use it – plagiarism and when – to do research, etc.
We have outlined tools such as ChatGPT, Bard, Firefly, and Harpa.ai.
Code of conduct – this outlines the ethics of AI use, current laws, and future laws. We also then have question considerations such as declaring the use of AI, legal implications, and weaponization.
Five general principles – this is voluntary compliance by regulators in conjunction with existing laws – safety, transparency, fairness, accountability, and contestability.
A final outline of top-level dos and don’ts – use transparently, use is to be unbiased, question and check, never share personal or identifiable details, and most importantly, report anything suspicious or not right.
Q: Would you mind elaborating on the "dos" and "don'ts" contained in your document?
TRANSPARENTLY use AI and do not claim things as original or unique.
DR: To be honest about when it has been used, you may want to include an asterisk in a blog to show where it may have been used to provide examples or, if questioned by a journalist, customer, etc., share when and where AI had been used and for what purpose.
FAIRLY use AI in any decision-making processes and remain unbiased.
DR: If using AI in a decision-making process, ensuring that it is unbiased is incredibly important. You may consult with AI to make sense of insights or reports, but you ensure that the responses are unbiased.
QUESTION AND CHECK facts, claims, source origin, and intellectual property, use AI and edit.
DR: In its early stages, ChatGPT had no knowledge past 2021, and some platforms like Bard were known for creating “hallucinations.” It’s always important to check facts and sources to ensure nothing is being plagiarised or generally incorrect information.
NEVER share personal or identifiable customer or business information with AI tools.
DR: We don’t know the full extent of AI usage yet; therefore, our topmost rule is to never provide it with identifiable data, especially that of our customers. We can use AI to help us sort data or identify patterns, before providing the data, we ensure that all identifiable data is removed.
REPORT anything that may seem suspicious or not quite right.
DR: Reporting on AI is super important, too. If something feels suspicious or not quite right, it's always best to flag it and follow up; it can help our business as well as others. For this reason, at the moment we only have certain AI tools available to certain team members whilst we trial and get to know the tools better.
Q: What dangers do enterprises face for not establishing an AI policy?
DR: An AI policy gives you a sense of where to go, what to do, and how to use AI. An AI policy ensures that employees know how they can do their job more effectively by identifying tools that can save them time. On the other hand, it’s there if a problem does arise, they know what to do and who to report it to.
An AI policy also protects us in the sense of copyright and plagiarism. As a business that produces a lot of visual and written content, we stick by strict policy regulations to ensure our work is original and not a carbon copy.
Q: What are some of the best practices for setting up policy templates?
DR: We took into consideration all the AI tools we’re currently using as a brand, as there are a few, and then dissected each platform from there. We’re still in our learning stage. We need to nurture AI with prompts and information to help them understand and perform tasks better.
We just make sure to avoid ever using any personal or identifiable information when using these platforms. We’ve created a working document with date stamps, as AI is evolving so quickly that our policies also need to be able to keep up.
The document is visible and accessible to all of our team to make sure everyone has eyes on it and knows what is expected. We encourage our team to feedback any concerns or highlights, AI policy writing is continuous while we’re still learning.
Q: Let's discuss the intersection of AI ethics and marketing since that's the focus of the newsletter. In what ways do you see an intersection? Are there policy tenets that apply specifically to marketing as opposed to other areas?
DR: Where other teams may use AI for automation for small tasks or identifying data, we use it daily for the nitty gritty of our jobs. From image creation to research, creating a blog template to providing alternative titles, we use it for so many different things. That’s why, as a marketing team, it is important that our employees adhere to our guidelines.
Q: Any further thoughts you care to share? The floor is yours!
DR: Again, it means that we’re transparent! At instantprint, we've fully embraced AI across various roles within the Marketing & eCommerce department. However, my main goal for the team has always been to maintain the production of high-quality work, even with AI assistance.
Interestingly enough, we're now using AI to assess and gauge the authenticity and "human" element of our content. It's quite ironic, but it allows us to ensure that our work strikes the right balance between AI-driven efficiency and genuine human touch.
We possess a grading matrix that was created in the past, which we have adjusted to incorporate AI detection. As a result, we have achieved a harmonious perspective on content that yields optimal outcomes.
By leveraging this matrix, we streamline our processes, gain inspiration, and establish foundational levels of structure. However, we understand that our audience appreciates the human touch and expertise, so we complement the AI-generated content with human experience to provide the quality they adore.
How would you respond to Robinson’s remarks? Should every business have an AI policy or ethical code in place? Leave a comment and let us know your thoughts.
As a premium subscriber, you know the value AI Marketing Ethics Digest provides its readers. Why not extend that value to a friend or colleague by providing a gift subscription? The cost is minimal — just $5 per month — less than a coffee at Starbucks.