A gray tsunami of algorithmic ageism

Much has been written about implicit bias in algorithms ranging from disparities between races and genders in facial recognition to misdiagnosis on medical apps and devices.  

According to the Pew Research Center, there are currently roughly 62 million adults ages 65 and older living in the U.S., accounting for 18% of the population. By 2054, 84 million adults ages 65 and older will make up an estimated 23% of the population. Even as the 65-and-older population continues to grow over the next 30 years, those in their 100s are projected to roughly double as a percentage of that age group.  

In the spirit of journalistic full disclosure, I fall into that over-70 demographic.  

So what could possibly go wrong with the representation of the elderly population across the massive universes that AI platforms (generative or otherwise) scour for output? Even more concerning is how these platforms will accommodate the unique informational and emotional needs of 84 million seniors in a digital setting.  

Most of the major fails with AI-driven platforms have stemmed from an incomplete universe from which insights are drawn. Many of these are related to race, gender, and social determinants of health being underrepresented, producing a skewed result and in many cases false positives and negatives. A speaker at a recent CHIME event creatively oversimplified it when he said, “If you only have an AI universe that looks like the Mona Lisa, your results will look like the Mona Lisa.”  

So to what extent is “father time” a factor in algorithmic bias? 

The answers are not that complicated, which makes the lack of solutions even more perplexing. One could argue that the market opportunity for many vendors is limited.  

First, as mentioned above, senior universes are underserved as it relates to many of the new trends in healthcare digital transformation. Yes, they are a huge part of the population that is treated for disease and other healthcare conditions. Yet, their relationship with the latest forms of health literacy content distribution is dwarfed by millennial and Gen X demographics.  

Yes, boomers and even their parents have embraced social media as a means of staying connected. But as these platforms rapidly accelerate, the magnitude of the impact on digital immigrants and economically disadvantaged digital have-nots increases disproportionately. 

Next, in a world where ChatGPT is in some cases years behind in its content curation the real world is operating at a digital warp speed. Think about how the world has changed since the 2022 cutoff date of the ChatGPT content grab. That could mean that two years of COVID-19 long hauler data for the senior demographic could yet be curated. The generative algorithms are just catching up with such breakthroughs as GLP-1 weight and diabetes management drugs which could have a significant impact on the senior segment.  

Also, many seniors are at the mercy of caretakers or their children in ensuring that their medical data is properly recorded in both structured and especially unstructured formats.  

Even in the healthcare universe as a whole, nearly 80% of patient data is unstructured and unable to be documented in EMRs with any degree of accuracy. But considering that many of the new applications and platforms are designed to reduce the epidemic of loneliness in senior demographics, this conversational voice-enabled data is far more important than most text-based structured data captured by humans or ambient technologies.  

Lastly, and perhaps most importantly, even if the senior universe is reliably included in the algorithmic spidering, is there a representation of this senior demographic mentality in the AI data scientist community to make sense of the data? One could argue that there need not be 70-year-olds on the dev teams for these generative AI products.  

But one could possibly argue that not having them leading collaborative teams is in itself the crux of ageism in digital transformation.