Content design and AI - Part 2: Experiments in user-centred design

Disclaimer: this blog references a pretend service. This is a purely fictional service and has been created for strictly illustrative purposes only. 

As I mentioned in the last blog, part two will explore an experiment I designed to see how AI handled some practical and true-to-life user-centred design.

I created a pretend large-scale service and set up a scenario whereby a multi-disciplinary team needed input from a content designer to help solve a challenge.

The pretend service and the associated challenges

This fictional service aims to essentially let patients book different types of blood tests once they have seen their GP and been given a referral. 

This service came about after a successful discovery concluded that the existing process for booking blood tests in England and Wales could be:

• A burden on existing NHS systems
• Too slow at handling the end-to-end processes for patients
• Hard to find within a busy digital healthcare ecosystem 
• Misaligned to emerging and changing policy requirements 

The service was mid-way through its Alpha phase and had a multi-disciplinary team working on it. Due to budget constraints and a lack of availability – one of the roles missing was a content designer. The team had a user researcher but they were only on the project part-time and had not managed to facilitate any accessibility research.

After the team had its mock service assessment ahead of private beta- it was recommended that they try to bring a content designer in as soon as possible to help iterate and align the service to better meet user needs and adhere more closely to the GOV.UK guidance on good content design. 

It was identified during the mock assessment that the two highest risk areas for a content designer to work on first were the service’s:

• Name
• Start page

The multi-disciplinary team would need a few different iterations of the proposed solutions based on evolving requirements – as is typical when working in fast-paced digital environments – and there would be a few specific criteria that the content would have to adhere to.

How the experiment will work 

I will be asking ChatGPT 4 to complete a series of tasks and iterations to help meet the team’s needs; ChaptGPT 4 will be acting as the content designer for the service. In this blog we will only be focusing on the work done for the service name. 

I will then peer review the responses and answers and give my feedback, commentary and insights as a user-centred design practitioner. 

Using AI to help with the naming of a service 

The first thing to do was to help improve the service name. So, I said to ChatGPT4: 

‘’We are working on a digital service that lets members of the public in England and Wales book a blood test for themself or someone they care for. The current name of the service is: Blood Test by formal GP Referral Only’’.

This was a name created by our policy team and was implemented before a content designer and a full-time user researcher were involved with the service.

The limited user research we have so far shows that people found this name to be a bit confusing, especially those with English as a second language. We have an opportunity now to explore re-naming the service along with the associated research in anticipation of moving into private beta.

How might we iterate this service name to make it better?

I kept the ask high level at this point so I was a bit cruel – but it is not uncommon for stakeholders to seek expertise from their designers as to how they can generally improve upon a situation before always having the luxury of detailed research insights.

Chat GPT4’s response 

To improve the name of your service, the goal should be clarity, simplicity, and inclusivity, especially for those who may not be familiar with formal medical or English-specific terminology. Here are some key considerations for renaming:

1. Use Clear and Simple Language:

• Avoid complex terms like “formal GP referral” that might confuse users, especially those with limited English proficiency. Use more familiar and universally understood terms.
• Focus on what the service does: booking a blood test.

2. Highlight Who It’s For:

• Include language that makes it clear who the service is for—people booking for themselves or someone else.

3. Be Direct and Concise:

• Keep the name short enough to be immediately recognizable, but descriptive enough to explain the service.

Safe to say I was reasonably impressed with this.

This is good guidance and responds directly to the ask.

It talks about keeping the language simple and removing complex terms which could confuse users.

It says to focus on what the service does, which is booking a blood test; this verb-led instruction is a really good steer and generally is solid advice that content designers would champion.

One I would lean into a little less is highlighting who it is for – a good service name will do this intrinsically but in fairness the rationale it gives is helpful especially if you are not from a UCD background but looking to learn more. 

Now, I specifically asked Chat GPT4 how it would recommend iterating the service name. I was hoping for guidance but it did one better and gave some recommendations of new service names to consider.

Suggested Name Iterations:

• “Blood Test Booking Service”

1. 1. • Simple and to the point, with no mention of referrals, focusing on the main action—booking a blood test.

• “Book a Blood Test (GP Referral Required)”

1. 1. • This iteration mentions the GP referral but in a simpler, more direct way.

• “GP Referral Blood Test Booking”

1. 1. • Keeps the GP referral mention upfront, but still focuses on booking the blood test.

• “Blood Test Booking for You or Someone Else”

1. 1. • Emphasises the feature of booking for another person, which could be important for carers.

• “GP-Approved Blood Test Service”

1. • Makes the process of needing a referral clear, but in a friendlier tone.

The first thing to note is that I believe all 5 of the names it gave are better than the control. Granted, I made the control name especially bad on purpose to help illustrate a point.

Of these 5 names, some are stronger than others and I would not be satisfied at this point to stop the ideation phase – but given this answer was provided in seconds these are good candidate names that form the basis of an iterative period.

The team’s response

These names were shared with the part-time user researcher for a light peer review. The researcher liked them and felt they were a step in the right direction for the name of the service.

That said,  they did want to see some further iterations which explicitly considered the GOV.UK guidance on naming a service. The researcher had also read a good blog by some former HMRC colleagues about a time they had to rename a service and wondered if any tips could be taken from their experience.  

So the team went back to Chat GPT4 and said:

Thank you Chat GPT4 for the help – what you shared with us was really useful. The team would like to see a few iterations of what you have done.

Can you expand on what you said but consider the following two articles in your response:

https://www.gov.uk/service-manual/design/naming-your-service

https://designnotes.blog.gov.uk/2024/05/08/a-user-centred-approach-to-renaming-a-service/

It is important that we adhere to the GOV.UK naming conventions as this was called out in our mock assessment. 

Thank you

Chat GPT4’s response 

You’re welcome! Based on the GOV.UK service manual’s naming conventions and the article on user-centred renaming, here’s a refined approach to iterating the name of your service while adhering to these principles:

1. “Get a Blood Test (with GP Referral)”

• Rationale: Starting with “Get” emphasises the outcome the user is looking for. Adding “with GP Referral” makes it clear that a referral is needed, but in plain, straightforward language.

2. “Book Your GP-Approved Blood Test”

• Rationale: This version highlights that the test is “GP-approved,” simplifying the concept of a referral and aligning with GOV.UK’s preference for common terms. “Your” adds a personal touch, aligning with GOV.UK’s user-centred language approach.

3. “Book a GP-Requested Blood Test”

• Rationale: Uses “GP-requested” instead of “referral” to simplify language. This follows the GOV.UK principle of using words users are more likely to understand without formal medical jargon.

4. “Book a Blood Test (GP Referral Required)”

• Rationale: Keeping it simple and clear, this version sticks to GOV.UK’s guidance to state exactly what the user is doing while noting the GP referral requirement in parentheses, making the name more accessible.

I wondered to what extent it would attempt to iterate the initial set of names it gave. Having seen the response I was quite impressed; comparing the names side by side – as someone who has renamed several citizen-facing services – I can see the nuance and the subtle changes that I would want to see.

Each new iteration was now verb-led whereas the initial set only had one verb-led name.

Comparing the initial recommendations with the second iteration is fascinating. The second iteration is more structured, with subtle but meaningful improvements. One constant is the recommendation ‘Book a Blood Test (GP Referral Required)’, which appears in both—clearly a strong suggestion. I liked it too, though I wondered how it might look without the parentheses.

The second iteration also shows progress in framing the names. For instance, while ‘GP-requested’ feels a bit clunky, it introduces a key concept: all blood tests must be GP-requested. Given the context and criteria, that’s a reasonable inference for the LLM to make at this stage.

Team response

Thanks again Chat GPT4 – these are really positive iterations which align more closely with GOV.UK guidance. Great to see they are all verb-led and that the referenced articles were factored in as part of your process. 

We would like to take some of these candidate service names into testing with users; our researcher wants to take in 2 names for testing. 

I’ve chosen 2 of the names your recommended in your latest iteration here with some further changes that the rest of the team want to see for each name::

• “Book a Blood Test (GP Referral required)”

Can you keep this name mostly the same but iterate it so it does not contain any parenthesis and removes the word ‘required’

• Book Your GP-Approved Blood Test

Can you keep this name mostly the same but find a way to replace or remove the possessive adjective ‘your’

Chat GPT4’s response 

Name Iterations:

1. “Book a Blood Test with GP Referral”
   • Rationale: Removing the parentheses and the word “required” simplifies the name while keeping the key information intact. This is still direct and easy to understand, aligning with GOV.UK guidance.
2. “Book a GP-Approved Blood Test”
   • Rationale: By removing “your,” this version remains concise and focuses on the action, making it more neutral without losing clarity.

Now, these are much better. 

There’s still room for improvement, but given how quickly it incorporated specific feedback, the results are impressive. Each iteration includes a rationale referencing the tweaks I suggested, with commentary on how these might enhance the user experience.

The service team now has two strong candidate names for testing—both far better, by content design standards, than the original ‘Blood Test by GP Referral Only Service’ (a truly dreadful name).

ChatGPT4 also highlights that a service name, while critical, isn’t the only way to guide users. It suggests that the referral requirement could be clarified within the description or user journey instead—a smart consideration that aligns with good service naming practices. While this would need testing, it’s a valuable insight and exactly what I’d encourage teams to explore, particularly for names with complex needs or policy constraints. Top marks, ChatGPT4.

So, what does AI mean for content designers?

Unfortunately, we’ll have to end on a cliffhanger until next week’s blog, when I’ll be going over my findings, observations and conclusions.