Motivation

Studies show that closed-loop (CL) systems can successfully support diabetes type 1 management and achieve better clinical outcomes than comparable treatment options. Such systems include many components, such as continuous glucose monitoring, insulin pump and a device that performs calculations such as insulin dosing.

"It’s hard to give up control. I think about my diabetes constantly. I make a million decisions a day. So it’s hard to let that go."
Participant in the study by Tanenbaum et al. (2020)

While the successful use of such a system means a great benefit to the quality of life of those affected, incorrect insulin dosing can have life-threatening consequences. Accordingly, it is important that users of such a system have complete confidence in the automated form of therapy. The goal of this project is to establish such trust through traceability of system behavior.

Research

To be able to dive deeper into the context, we have worked intensively with the LoopInsight1 (LT1) simulator. LT1 is an open source project from the University of Applied Sciences Ulm, which simulates the interaction between a person with type 1 diabetes and an AID system. The simulator of LT1 forms the algorithmic basis for this project. Therefore, we took the functional basis of this simulator and adapted the user interface to our findings, e.g. to strengthen the promotion of trust in the system.

Challenges

We identified the following challenges in the development and design of AID systems:

Interviews

To learn more about the context, we conducted interviews with three people with type 1 diabetes. The goals of the interviews were to find out what data are necessary for diabetes management and what problems or challenges are encountered in this process. These interviews were also a content entry point to learn more about the context of use.

"At the moment, I have to trust the system blindly and hope that it makes the right decision. Of course, it's also very difficult in the first few weeks to completely relinquish this control."
Interviewee who currently uses a closed-loop system

As a first step in building trust in closed-loop systems, interviewees expressed a desire to be able to test the functionalities without effects on their own bodies. Existing systems are far too non-transparent and do not offer any explanations, according to the interviewees.
In addition, the desire was expressed to be able to use the simulation spontaneously and on the go, in order to be able to play through spontaneous activities for safety's sake.

Requirements

We were able to derive some requirements for the design of the simulator from the interviews:

• Reduction and simplification of parameters to set the simulator
• Provide insights into the calculation process
• Communicate capabilities and limitations of the system
• In a wide variety of everyday situations, the question could arise as to how the system would react. Therefore, the simulator should be accessible on the go with mobile devices.
• Offer scenario presets that represent a specific situation that affect blood glucose

In particular, these requirements are intended to promote the level of trust in the system through familiarization with its functionalities and limitations.

Design

The result of this project is a mobile AID simulation app based on the LoopInsight1 algorithm. As a basic requirement for the design, we focused on usability for a broad target group. Therefore, we have reduced the adjustable parameters to the most necessary and extended them by an expert mode with further setting options.

The use of Google's Material Design 3 guideline is intended to help users recognize familiar interactions and reduce the required learning of functionality.

Dealing with the first-failure effect

In order to deal with the first-failure effect, we allow the users to actively observe challenging scenarios (e.g. through several preset scenarios). This includes not only communicating the scenarios that can be covered by the system, but also functional limitations to avoid overreliance.

Ability through traceability

In order to gain an understanding of the system behavior, users should play around with the system themselves as interactively as possible. For this purpose, we have introduced two different types of explanations:

• Local explanations — Users can check out extreme points on the curve and get insights on the expression of the data at that point
• Global explanations — An interactive visualization of the data flow conveys the influence of the inputs on the calculation process. Users can thus track step by step which changes certain inputs result in.