Diabetes self-management is behavioral

The personalization pipeline shipped before the consumer web — as Diabetes Talk, an interactive telephone counseling program launched under the Eli Lilly brand for people facing the move from oral medication to insulin. Callers worked through a counseling session in the encoded voice of psychologist Richard Rubin — the ADA’s Outstanding Educator in Diabetes for 1997 — and received a 16-page booklet generated from their individual concerns, plus a personalized summary designed to be brought to their doctor. The launch brochure below is the original printed object, die-cut tabs and all.

The artifact · Eli Lilly launch brochure, 1997 · flip it by its own tabs

Cover 3 Questions Diabetes Talk The Program Calling Guide

The counselling logic moved to the browser, and for the first time the program could answer a person on the same day they spoke to it. Rubin appeared on video inside the page. Validated psychometric items were rendered as clickable response buttons, and a confidence gauge swung as each answer landed — research instruments turned into learning experiences. A scale you complete became a scale that answers you back.

Over the next four years it grew into a whole program rather than a set of sessions: a multimedia health and behavior assessment, Advisor Sessions and Self-Care Clinics built around real people telling their own stories, a progress tracker, a scrapbook the user filled themselves, and a care planner meant to make the patient an active participant in their own clinical decisions rather than the recipient of a plan.

The delivery was its own invention. Video and audio lived on a CD-ROM in the box and were unlocked only by a key from the server — broadband media in a dial-up world. The pilot shipped with a home HbA1c test kit in the same carton, sponsored by Roche and run through Humana, so the program could speak to a number that actually belonged to the person reading it. What follows is the real build, Netscape chrome and all.

The approach went global, and changed audience. A 2003 Novo Nordisk engagement produced the DAWN Interactive CD (2004), which took the DAWN study’s findings on the psychosocial burden of diabetes — the distress, the fear of insulin, the things a clinician rarely has time to ask about — and turned them into interactive training for the providers themselves.

For the first time the program was teaching the clinician rather than the patient, developed with named behavioral experts and shipped to national affiliates to translate and adapt for their own countries. It was the furthest this work ever traveled and the least it ever learned: discs went out by the thousand, and nothing came back but opinions. Guidance that changed after pressing stayed wrong on every disc already in the field.

Deployed as part of a comprehensive disease management and clinical decision support system for a diabetes clinic, integrating chronic care registry, point-of-care decision support, remote data capture, and HL7 interoperability into a team-based care environment.

What once required IVR scripts and hand-built rule engines now runs on conversational AI: truly adaptive coaching dialogue, natural language understanding, and the same encoded clinical wisdom — six generations on, delivered by an avatar care team. The behavioral science stays the same. The technology finally matches the ambition. This generation’s artifact is a working prototype you can use right now — take the same kind of assessment you just answered on the 1998 screen, and watch what six generations of refinement did with it.

The current generation · 2026

A conversation that remembers, and changes its mind

Six generations of this program measured people and told them things. This one talks with them — one question at a time, branching on what the answer actually means rather than on which button was pressed, carrying what it learns from one topic into the next.

The behavioral science is unchanged. What changed is that the machine can finally hold a conversation, so the method no longer has to be flattened to fit the delivery.

01 · How it opens

It asks before it tells

One question per screen, never a battery of them. Something of value comes back before very much is asked for. Every answer changes what comes next.

Declining is a legitimate answer — and a permanent one. Anything somebody passes on is never asked about a second time, which is the difference between a conversation and an interrogation.

An assessment item asked at its moment of relevance is not a question. It is the conversation.

02 · What it watches for

It notices the turn

The moment worth catching is when somebody moves from I can’t toward maybe I could. It rarely announces itself; it arrives mid-sentence, in a change of tense or a softening.

The program follows readiness rather than a script, and when the ground shifts it offers to change direction rather than simply doing it. With permission, and only once.

A door opening, not a trophy — the visual language for the moment somebody turns

03 · What there is to talk about

Nine doors, entered in any order

Understanding

Blood sugar

Food

Medications

Moving, stress & rest

The emotional side

Your people

Looking after the rest of you

Your plan

Nine doors, all of them made — entered in whatever order the conversation takes, not in the order they are listed here

04 · What you leave with

Cards you chose to keep

A takeaway worth having is one you selected, not one you were issued. Anything that lands can be kept — a reframe, a technique, a sentence to use on somebody who keeps commenting on your plate — and it goes into a collection that is yours.

This is the oldest idea in the program. In 2000 it was called the Scrapbook, and the brochure described it as a place that is all yours, to put text and pictures. Twenty-six years later that is still the best description of it anyone has written.

Card faces — the visual carries the attention; the teaching sits behind Learn more

05 · What the clinician gets

Two documents, and one of them is optional

At the end there is a document for the person and, if they choose to share it, one for their clinician — what matters to them, what is in the way, what they have decided to try.

Sharing is always the person’s decision, asked at the end rather than assumed at the start. A summary that arrives without consent is surveillance, however useful its contents.

06 · Where the words come from

Nothing reaches a patient by being found

More than six hundred documents sit behind this program — thirty years of curricula, assessments, expert contributions and deployment. It is the asset and it is the hazard, because guidance ages and licenses vary.

So the archive is never machine-readable. Not indexed, not embedded, not searchable. Material is read by hand, rewritten, documented with its source and its license, and only then published. Presence in the program is itself the evidence that somebody put it there deliberately.

The measure of a content gate is not what it lets through. It is what it stops.

07 · What it needs now

Three things would move this further than anything else

Clinical partnersWilling to put their judgment against ours and review the program properly People living with diabetesA small number, prepared to use it honestly and say where it falls short A care settingCurious what a behavioral program looks like built from thirty years of practice rather than a content library

With the technology no longer the limiting factor, the method is the only thing left that can fail. It has had twenty-eight years to get ready.

‹ Back Next ›