By Claire Lynch
When artificial intelligence began moving rapidly into everyday life, Tobey Dichter saw both possibility and risk.
As founder and CEO of Generations on Line, an organization that has spent decades helping older adults build confidence with technology, Dichter was already thinking deeply about loneliness, social isolation and the digital divide. Then, almost coincidentally, she attended a conference where Microsoft representatives were describing how quickly artificial intelligence would develop.
Her immediate reaction was not simply, How can older adults use this?
It was: Where do we begin? Is it legitimate? And how do we make sure we don’t cause more harm than good?
Those questions became the starting point for an ambitious Innovative Project Support initiative funded by The Sarah Ralston Foundation: developing an age-friendly way for older adults—particularly those with limited digital experience—to understand and use conversational artificial intelligence.
“SRF’s Innovative Project Support program is specifically designed to help organizations develop, test or extend new approaches to serving vulnerable and underserved older adults,” said Heather Finnegan, Executive Director. “For Generations on Line, that mandate offered permission not simply to launch a new program, but to experiment.”
The grant was not intended to underwrite a finished product. It was seed funding for an idea that still needed to be tested, challenged and, if necessary, rebuilt.
“We knew we had to dive in deeper,” Dichter recalled. “We didn’t want to get into anything that would create more damage.”
Asking Older Adults First
Rather than assuming what older adults wanted from AI,
Generations on Line began by asking them.
Generations on Line began by asking them.The organization conducted five focus groups at sites serving older adults who might face financial, social or technological barriers: John C. Anderson Apartments, affordable housing for LGBTQ+ older adults; the William Way LGBT Community Center; Germantown Apartments, part of NewCourtland’s affordable senior housing; and two Federation Housing communities for low-income older adults, Reider House and Goldsteins’ Rosenberg Residence in Northeast Philadelphia.
The first surprise was that participants were far more receptive to artificial intelligence than the team had anticipated.
According to Dichter, 67 percent of participants said they felt somewhat or very positive about AI.
That gave Generations on Line a place to begin—but not a finished product.
The organization developed what would eventually become IMAComputer.org, a training program and conversational AI experience created specifically for older adults. It included introductory videos, guidance about safety, explanations of generative AI and an interface designed around the ways older adults actually approach technology.
The development process was highly iterative. The team conducted pre- and post-testing, returned to participants with prototypes, revised its training around two-factor authentication, and continued changing both the curriculum and the underlying architecture of the interface—what was taught, how it was taught and how interactive the experience could become.
At one point, Dichter said, the team essentially decided to “tear it up and start over.”
That, in many ways, was the innovation.
Designing for a Different Starting Point
Generations on Line has long understood that many older adults enter the digital world from a fundamentally different starting point than younger users.
Technical vocabulary that developers take for granted may be unfamiliar. Interfaces can move too quickly. A system that interrupts or expects an immediate response may feel frustrating rather than intuitive.
So the team built IMA around older adults rather than asking older adults to adapt to conventional AI.
Early testing explored dozens of possible conversation prompts before the organization narrowed them to a small number that users found approachable. They were: “There’s something from the past I’d like to remember,” “I want to create something fun,” “I could use some advice,” and “I need to vent with no judgment.”
That last one proved especially popular during pilot interviews.
Another lesson came from the way people interacted with the system. Many initially typed their responses because that was how they believed a computer was supposed to work. Once participants realized they could simply talk, the experience became much more natural.
Generations on Line now plans to emphasize voice interaction even more prominently.
The tool can also help with very practical tasks. An older adult might ask for ideas for a birthday gift, create a meal plan, develop a shopping list, or adapt meals around allergies. But the goal is not simply to reproduce a general-purpose chatbot.
IMA was intentionally configured to respond in ways that better suit older users: allowing more time, avoiding unnecessary interruption and repeatedly reinforcing safe use.
When asked how she works differently from a regular ChatGPT conversation, IMA herself offered a succinct explanation:
“The simplest way I’d put it is: your regular ChatGPT is a general-purpose assistant; IMA is ChatGPT that has been given a very specific job, audience, and manner of speaking. In this case, IMA has been instructed to behave especially well with older adults who may feel uncertain about technology. So compared with your ordinary ChatGPT conversation, I’m deliberately more likely to explain things in plain language, break procedures into small steps, avoid technical jargon, reassure someone who is worried about making a mistake, and call attention to scams, privacy, suspicious links, and other internet-safety issues.”
Safety Isn't an Add-On
As the project evolved, safety became even more central.
Generations on Line tested how the system responded when users raised sensitive issues, including depression and suicidal thoughts. The organization wanted to make certain that IMA would recognize when a conversation had moved beyond what an AI tool should handle and direct the individual toward appropriate human or medical support.
That distinction matters especially as generative AI becomes increasingly persuasive and conversational.
Dichter said skepticism among participants was not something the organization tried to eliminate.
“The skepticism is warranted,” she said.
Instead, the training teaches people both what AI can do and where its limitations and dangers lie.
Privacy was also a deliberate design consideration. According to Dichter, Generations on Line and its Board chose to structure IMA so that users’ conversations remain within the IMA environment rather than becoming part of the broader ChatGPT system—a choice the organization made despite less expensive alternatives. The goal, she said, was to give older adults greater confidence as they practiced using AI.
And those safety considerations continue to evolve—which means the program has to evolve as well. Dichter said one of the team’s next priorities is strengthening safety education as new AI risks emerge.
A Door Opens Through Language
One of the most significant developments came in the project’s second year.As Generations on Line continued refining the system, it became clear that conversational AI could help address another longstanding digital barrier: language.
If a participant speaks to IMA in Spanish, the system can respond in Spanish. The same is true for Russian and other supported languages.
For Generations on Line, that capability opened an entirely new avenue for reaching older adults who had often been excluded from digital-skills programming because of language barriers.
Dichter recalled one training session conducted in Spanish in which a participant began skeptical of the technology. Midway through the session, the woman suddenly blurted out:
“I need this in my life.”
It was the kind of reaction the team had hoped for—not enthusiasm for technology simply because it was new, but recognition that the technology could actually be useful.
From Curiosity to 94 Percent Interest
The numbers emerging from the project suggest that reaction was not unusual.Interest in AI rose from 67 percent before participation to 94 percent afterward, according to Generations on Line’s post-program surveys. Eighty-seven percent said they would recommend the experience to a friend.
After launching at partner sites, IMA has now reached approximately 334 beta testers.
The pilot expanded beyond the original focus-group locations to include the Philadelphia Senior Center and other community partners, giving the Generations on Line team additional opportunities to observe how older adults interacted with both the training and the technology.
Those numbers are still relatively modest, but the structure of the program points toward much broader reach.
Because IMA does not require a fixed physical program site or a one-to-one staffing model, the potential capacity is enormous. Generations on Line envisions expanding access throughout Philadelphia through libraries, senior centers and other community settings, as well as eventually reaching people directly.
“What we have created has infinite capacity,” Dichter said.
And because the program is free, scale does not necessarily have to mean creating another financial barrier for the older adults it was designed to reach.
Generations on Line has also developed a coaching guide and other training materials to help community organizations support older adults as they learn to use IMA, extending the project beyond the technology itself.
What Seed Funding Made Possible
In 2025, after the project had begun demonstrating its potential, Generations on Line received additional support from the Philadelphia Foundation. The organization is now pursuing further funding to continue refining and expanding the program.
But Dichter is quick to point out that relatively few funders were willing to take a chance when IMA was still simply an idea.
The Sarah Ralston Foundation was.
“We were so excited when we saw the word ‘innovative,’” she said.
That willingness to fund experimentation gave a small organization with nimble in-house technology capacity room to test an idea, listen to older adults, discover what wasn’t working—and rebuild it.
In that sense, the grant bought something more valuable than a finished product: it bought the organization room to learn.
Only after the project demonstrated that older adults wanted the technology, could use it and found it valuable did Generations on Line have the evidence needed to pursue larger-scale support.
What began with a relatively small investment in an untested idea has now grown into a program with hundreds of beta testers and the potential to reach older adults far beyond its original pilot.
The result is not simply another AI chatbot.
It is an attempt to ensure that older adults are not passive observers as one of the most transformative technologies of their lifetime develops around them.
And perhaps the most innovative part of the project has been the willingness to keep asking the same question that began it:
How do we make this useful without making it harmful?
Visit the ImaComputer introduction on YouTube: