A 7 week individual project at AHO

Collaborator

Tietoevry - AI Insight Team

What did I do?

UX Design, Illustration and storytelling, Mapping User Experience, Prototyping (Figma) & User Testing


Exploring human-AI interaction patterns for public sector

This was a student internship project as part of the Master in Design program at AHO in 2023. The task was to explore AI as a support tool in the archiving program Public 360 for the AI Insight Team at Tietoevry.

Nominated for AHO Works!

This project was nominated by BEKK for the Prize for Interaction Design in 2023 at the annual AHO Works Exhibition.


Background

The Collaborator

TietoEvry is one of Norways biggest IT companies. They made a software called Public 360. It’s primarily for archiving documents, and some municipalities in Norway use it.

Politics play a role..

That’s because we have something called
Offentlighetsloven in Norway. It states that the country is transparent, and most documents have to be accessible for the public. But at the same time, there could be a lot of sensitive information in these documents. Like your bank account number or address. That’s why we have something called Personvernloven. It states that the sensitive information that could say something about a person, business or other confidential stuff is to be withheld.

The Challenge

Public 360 as a software

We’re facing some problems. The roles in 360 are divided into “arkivar” and “saksbehandler”. The first are the experts that usually do the quality control (that everything is archived correctly), and the latter is everyone else. They both have the same interface.

The chosen user group

I chose to work with the group “everyone else”, because some of them cause a little bit of problems.

1. They usually forget how the routine works and mostly need assistance. And if they’re on their own, they don’t know they’re archiving incorrectly.

2. They also “just handle” cases, because they have a lot of other things to do, and the focus is just to get the document into the system.

3. And they often label the sensitive information as public when it should have been shielded.

The Non-Frequent User

We call this group the non-frequent user. They could access the program everything from once a year to a couple of times a month. They produce some of the documents that are put into public 360 everyday - including the sensitive information.

That’s both a major quality breach and safety risk. And every time something leaks through, they have to pay a fine.

Through interviews, I found out that most users don’t even know they’re archiving incorrectly.

Mapping User Experience

But If I want to figure out how to give support, I have to find where they typically struggle, so I had multiple meetings with my collaborators, analysed their program and also did user testing with a client of theirs.

The registration-forms had no visual aids, a lot of clicks and words that are hard to understand.

Overarching issue

In short - the user struggles with information overload. It’s really hard to prioritise to label the document as public or not (which this basically is all about).

Concept Development

I took it into three categories to develop Tone of Voice and functionalities -

The first (which I called intrusive) led me to a couple of issues - when is it too intrusive or demanding that it takes over the user.

The second (which I called preferences) is about the personality of the support, as I like to call it. What if you never want to hear from it again or think it’s annoying? What if you appreciate that it offers help when you struggle?

The third (which I called cognitive overload) is all about the information overload. How do you help a stressed user that has too much to look at and important tasks to do?

Prototyping and Testing

I took it to my collaborators and they pointed out they still wants the user to review the document manually, not just look at the findings, and to look for additional sensitive info if the document was indicated.

Trust - an important factor when working with AI

The action points are too focused and will give the user tunnel vision, especially when the AI could be wrong or not pick up all the info. They still want the user to read the document.

Playfullness

I steered the visual identity towards the TietoEvry logo with a hint of playfulness, using the Roboto font and visuals (or icons). And tried to simplify the content without loosing essential info.

Result

This concept became an AI support tool. It's an online extension compatible with the archiving software Public 360. The extension is meant to guide the non-frequent user through the archiving process, and to ultimately minimise the amount of sensitive data that is leaked to the public. My focus point was to map the issues and user experience of Public 360, and cover the user needs in the extension's interface. This project was in collaboration with TietoEvry, and was meant to spark a discussion of how the AI could be used to improve sensitive data management in the future.

Next
Next

Personal Projects