Designing the iLet: A Design Process in 5 Parts
User experience design and clinical integration for the Beta Bionics iLet automated insulin delivery system. This four-part series documents the design process, from initial kickoff to final interface decisions.
Part 1: Kickoff
By Sara Krugman | July 13, 2015
This article launches a four-part series examining the design process for the iLet, Boston University's Bionic Pancreas device. The piece describes the initial kickoff meeting and foundational design principles that guided the user interface development.
Design Challenge
Current CGM monitors and insulin pumps operate independently. While the automated logic seems simple—"sugar goes up, give insulin; sugar goes down, give sugar"—real-world implementation involves significant complexity including regulatory hurdles, safety margins, and technical constraints.
Foundational Approach
The team established four design tenets:
- Be reliable and transparent, show cause and effect
- Communicate respectfully, recognizing people with diabetes as experts
- Minimize user interaction where possible
- Design with safety as priority through intentional entry and customizable alarms
Control & Collaboration Model
Rather than full automation, the design philosophy emphasizes collaboration between user and device. As the author explains: "The interaction with the iLet™ is not about passing off everything. It is about collaborating in the management of blood sugar levels."
Functional vs. Experience Goals
Functional goals address device communication and status reporting. Experience goals include trustworthiness, transparency, simplicity, and user empowerment.
Part 2: Mental Models and User Research
By Sara Krugman | July 16, 2015
This post documents Tidepool's user research process for designing the bionic pancreas interface. Since the team couldn't participate in clinical trials, they conducted five interviews with trial participants who had worn the prototype for one week.
Research Methodology
The researchers used multiple techniques:
- Structured interviews asking about emotional experience, safety, trust, and mental model shifts
- Card sorting exercises to understand how users prioritized information differently with the bionic pancreas versus traditional pump therapy
- Paper prototypes to explore specific interactions (CGM disconnection, carb entry, target adjustments)
Key Findings
1. Learning Curve Required
One participant noted: "I trusted it, after I learned it...every time you have a first, it's not going to do well, but after that…" The device needed time to learn individual patterns.
2. Contextual Carb Entry
Users thought about carb entry based on their daily context rather than meal times or clock time—a significant shift in how they conceptualized diabetes management.
3. Information Hierarchy Shift
Participants emphasized that "you don't have to know all of that stuff—it's neat to see, but you don't have to know that." This indicated a fundamental change: data previously considered essential became supplementary with automation.
Mental Model Transformation
The research revealed a critical paradigm shift. With current insulin pumps, users focus on "How am I doing?" With the bionic pancreas, the primary question becomes "How is the device doing? Does it have everything it needs?"—reflecting how automation changes diabetes self-management.
Part 3: Interface Details
By Sara Krugman | July 20, 2015
This blog post examines design decisions for individual screens and interactions in the Bionic Pancreas (BP) interface, emphasizing a layered architecture that ranges from simple, glanceable information to deeper interface features.
Main Screen Architecture
The design team chose a map-based approach rather than a linear path. The interface organizes information around three fundamental questions, progressing from immediately accessible features to buried deeper options. This structure acknowledges that people with diabetes are knowledgeable experts in their own care.
Safety-Focused Design
Alarm Management
The team minimizes alert fatigue by requesting only necessary interactions. They incorporated "preparing" screens before insulin or glucagon delivery, offering a pause for cancellation.
Confirmation Alternatives
Rather than traditional confirmation screens (which create routine button-pressing), the design enables users to "undo" entries after submission, providing genuine safety without training habitual confirmation behavior.
Contextual Feedback
Icons displaying device status appear prominently, allowing effortless visual assessment.
Intentional Entry Design
Lower-impact carbohydrate entries reduce harm from errors, since the device continuously checks and corrects calculations.
Specific Interface Elements
Lock Screen
The lock screen prevents accidental inputs affecting therapy. The team evaluated interaction severity and added undo capabilities for higher-impact features.
Alarm Hierarchy
The system employs four escalating communication levels:
- All is fine (no action)
- Notifications (self-clearing)
- Alerts (acknowledgeable, snooze-able)
- Alarms (requiring action, non-dismissible)
Carbohydrate Entry
Rather than restricting entries to breakfast, lunch, and dinner, the revised interface uses a circular, cycle-based visual. This approach emphasizes metabolic timing over clock time, reflecting that the algorithm learns individual insulin sensitivity and carbohydrate response patterns.
Number Entry Methods
The team selected different input methods based on context: keypads for blood glucose entries (diverse, intentional inputs), and up-down buttons for weight adjustments (minor changes from existing values). The eInk screen's slow refresh rates eliminated scroll wheels, which would cause ghosting artifacts.
Design Philosophy
The overarching takeaway emphasizes that "simple doesn't come from simple"—achieving simplicity requires iterative design, real-world testing, and strategic decisions about what to exclude. The authors note that constraints prove essential to effective design and that perfect outcomes require releasing functional prototypes to understand actual user behavior.
Part 4: FAQ
By Sara Krugman | July 24, 2015
Display Technology
The team selected an e-ink display for its technical advantages and user experience benefits. As the author explains, "eInk displays are super thin and use very low power." The black-and-white format supports the device's character as a reliable, unassuming medical tool rather than a consumer gadget requiring flashy animations.
Power Efficiency
Color OLED screens demand unnecessary power for visual effects. The author shares a personal anecdote about an insulin meter that consumed battery power displaying an animated logo, leaving insufficient power for actual blood glucose testing—a frustrating design flaw the team wanted to avoid.
Physical Size
Patent protections on motor and drivetrain components influenced the device's dimensions. The screen size drove the overall form factor, with the understanding this represented version one and future iterations would be smaller.
Power Source Options
The team chose AA batteries to minimize external dependencies and supply chain risk. However, the author acknowledges tradeoffs: batteries offer flexibility for travelers, while rechargeable options appeal to users accustomed to managing multiple devices.
Typography & Tone
Using lowercase letters reflected intentional design philosophy. The approach conveys approachability—"lower case letters are less demanding, less assuming" than uppercase alternatives.
Team Composition
Success required complementary skills, clear communication, and effective project management to prevent scope creep and unnecessary debate.
Open Access
Design files and prototypes remain publicly available for review and implementation.