February 22, 2023 - Updated, priorities of the week

Back to GC Web Priorities Meetings - Réunions des Priorités Web du GC

Meeting information

Agenda

Updates / Mises à jour: DTO, PCO, PP, HC, other departments
Approach to removing RAP - Peter Smith
Using ChatGPT in the design process - David Pepin

DTO updates

DTO is available for questions concerning impending footer update
- dto.btn@tbs-sct.gc.ca
Heads of comms message for the footer update coming

Approach to removing Report a problem (RAP) - Peter Smith (DTO)

DTO and Principal Publisher are looking at retiring the Report a problem (RAP) pattern where it’s implemented as a form that doesn’t allow for write-in feedback.

Report a problem

The current hypothesis is that the community isn’t using the RAP data because it doesn’t include specific information about what’s wrong with a page. The current implementation offers only a checklist of possible issues with no free text write-in field.

Action item: To ensure that removing the pattern won’t pose an issue to anyone, please let DTO know if you are using the data for reporting or optimization work.

This isn’t the same as the new Page feedback tool that does include write-in feedback. It also doesn’t apply to departments who have implemented Report a problem with a customized form that does allow feedback.

Page feedback tool

This pattern can replace Report a problem on pages where you want to collect feedback. It isn’t meant to be deployed on all pages as you need to be able to review and act on feedback that comes in. If you’re interested in knowing more, you can set up a discussion with Lana Stewart.

Action item: We’d like to hear more about how people are using feedback they’re collecting to improve content. If you’d like to share what you’re doing, contact DTO about a spot on a future agenda.

ChatGPT possibilities in our design processes - David Pepin

Presentation: ChatGPT for #GCDigital: How this technology could impact our work

Note: This is not a set of vetted guidelines. It’s just notes on some experiments using ChatGPT as a way of starting some conversations about how it could impact our work in the future.

Discussion

ChatGPT operates in French and can easily translate other languages.

Chatbots and virtual assistants still require well-designed content as a base, so priority should still be good web content.

ChatGPT is meant to generate probable answers but needs to be monitored. It can’t be autonomous as a potential tool, for example, for clients submitting accessibility tests.

Hasn’t been vetted for dis/misinformation
Pulls in information indiscriminately and can’t distinguish between what’s true and what isn’t
Very confident in its answers, even when they are totally wrong
Not optimized for generating true text, just plausible-sounding text

Important ethical questions:

It could be helpful to use ChatGPT to analyze user feedback, but is it ethical to feed comments into an AI without being transparent about that?
If GC’s role is to provide answers on its services, is it ethical to implement ChatGPT when the help it provides does not reveal the same answers all the time to all users?

It is an unreliable communicator. The possibility of more interactive pages with wizards and calculators is exciting. People want answers, not information. But there can be a mismatch between the information people want and the reality of what is actually available - like the desire for vaccines in the early days of covid - there wasn't any vaccine, but people wanted it. The way it works right now ChatGPT would tell people what they want to hear rather than the truth.

There may be interesting implications for search. ChatGPT will certainly disrupt search. We may see AI-generated blogspam dominate search results.