The Ethics of using Artificial Intelligence to Create Developer Content
The need to keep developers up to date with the latest and greatest things going on in any technical ecosystem, from Open Source Software to Enterprise product offerings, requires content in many different forms. Videos, tutorials, webinars, documentation, event presentations, and many other conduits feed into the developer content machine and developing content that stays fresh and informative is a constant struggle.
Some schools of thought have turned an eye toward Artificial Intelligence as a way to consistently keep content moving forward. Opinions are divided as to the ethical nature of this technique. Here we will look at if and when it makes sense to use AI ethically to produce content geared toward the developer audience.
Fresh Content
When we consider fresh content for developers, we are referring to greenfield content. This need not mean something groundbreaking (though it can) but more content developed for the first time in a specific format or for the first time by the agency producing it.
Fresh content comes into play when a new product or project is developed, when a new technique has been developed, or when something that exists is being used a new or fundamentally different way. This can include things like releases with breaking changes, a pivot in what the software does, or applying a new feature in an unexpected way. Developers currently using the existing software or able to leverage it in the new way rely on content to ensure their skills are at par with what the application can do.
In the case of fresh content, AI should be used lightly. Generally, there will not be a great deal of existing content or information for an LLM or AI service to have been trained on, meaning there would likely be very little to do more than generate some templated content, lacking in specificity.
At this stage, it is likely better to avoid using AI to developer this content. Retaining an organizational tone and a writer’s voice here, along with making sure the content is correct and accurate, is paramount. Especially in the case of announcement or documentation for something fresh, the content should still be created by people with the technical understanding and the ability to educate. AI cannot help to retain that essential content at this point.
Updating Existing Content
Existing content is often a different story altogether. In this case, we have a piece that is out of date and likely needs less editing and fewer changes to bring our developer audience up to date. AI may be able to assist in this.
The point of how much to leverage AI on this is dependent on the content type. If we are simply updating existing blog posts or documentation, it may be beneficial to run the content through an LLM aware of the latest updates to the application or software. This can help us catch all points where changes need to be made.
It would be important here to consider if fresh content would not be better. If it’s really about a minor update for something relatively inconsequential, we can use AI. If it is a major version change or the level of complexity grows significantly, such as with and LLM going from a simple use to multi-modal functionality, it may be best to consider developing fresh content and leaving the existing content in place.
When it comes to larger scale changes to a tool or application, developer adoption often takes time. Immediate universal adoption is rare, but content for both the latest version and older versions is an excellent way to retain developers at various stages in their user journeys.
Video and Other Non-written Content
When it comes to video and non-written content the rules suddenly change. Generally, these feature a voice or even an image of a real person who is presenting the information and engaging with the audience, either in a pre-recorded way or in a live webinar.
Here is where the ethics of AI use is essential. Representation of a company, product, application, or organization is dependent on relatability. While AI has come a long way along the road of creating realistic images and content, it still falls short in the area of voice and human image consistency.
In the case of videos and webinars, using AI to create the content would be unethical, but there is still a place for it in the delivery of information. Transcripts and Captions created by AI and edited for consumption by humans is an excellent method for guaranteeing the content is inclusive, clear, and concise. Beyond using AI as a tool in visual content creation, it should probably be avoided at this point in time.
Event Presentations
In the case of event presentations we probably have the lightest touch when it comes to leveraging AI. A presentation at an event is meant to be impactful but also personal. It needs to include all the correct information along with the personal stamp of the speaker. Delivering something that comes across as mechanical or cold can often put developers off on the topic and could even cause backlash. Using AI to help in your understanding of a concept or technology is reasonable, though should always be backed by checking the human created literature for any nuance or edge cases. Like checking the side effects of a medication, it’s important to know how something functions in the context of how you are using it and how it applies to the things you are presenting. Given the bespoke nature of presentations for events, avoiding AI (unless that is the entire topic) as much as possible would be advised.
With the understanding of the maturity of current Artificial Intelligence tools and LLMs, treading lightly when creating content is the current recommendation. While it may become commonplace in the near future, leveraging tools for content creation in an ethical way is still some time away from becoming commonplace. Again, while it may great for minor tweaks, idea generation, static images (in moderation), and iterative updates, we still need to rely on content creators to ensure we carry a message that makes sense and is of value to developers.
