Augmenting Human Productivity With Artificial Intelligence
Given many of the challenges that we have today with obtaining high volumes of high-quality data to train advanced analytical models that drive artificial intelligence, we are really at the point of using artificial intelligence to augment human activity versus replacing it.
There is a term for this, Augment Intelligence which is also referred to as intelligence augmentation and cognitive augmentation. These terms are used to refer to analytical modes that complement and not replace human intelligence. These are descriptions that convey that the tools are helping humans become faster and smarter at performing tasks. The tools themselves are not technically different from what is considered artificial intelligence.
One could think of artificial intelligence as a continuum. As the analytical models are refined by producing many results that are either accepted or rejected, they become what we would call “smarter” or “more intelligent”. Generally, the process of “accepting” or “rejecting” the results is performed by humans. Sometimes, this entire process is conducted “in house” by trained experts prior to releasing a well defined commercial model into the market and other times models are released into an environment that provides imperfect results, but still provides some value to the user who are then willing to refine the results to obtain some level of value exchange between the results producer and the user. Thus involving users in refining the model so that it becomes more and more accurate or “intelligent”. There are many examples of this in the market.
Intelligent Humans Training Artificial Intelligence
One of the first ones is the CAPTCHA, you know that little image of garbled letters that many sites would use to determine if you were a person or a spambot before they would let you sign-up for a new internet service. Remember these:
Well, what you might not know is that the team of data scientists from Carnegie Mellon University led by Luis von Ahn that designed the CAPTCHA which stands for “Completely Automated Public Turing test to tell Computers and Humans Apart”, we’re getting us to perform a very important task, to translate nonsensical images into text for digitizing things like the archives of The New York Times and old books for Google. The value exchange for each of us to do the work was access to the content for which we were registering.
The collective impact of our efforts was that Google and others continually refined/ trained their analytical models to the point where the models were so good (accurate) at reading the nonsensical images that they no longer needed humans to do the work. Our collective efforts, in the beginning, resulted in a more seamless and quicker user experience to access valuable content while still protecting the website owner from annoying and potentially dangerous spammers.
Google, Amazon, and Netflix use Artificial Intelligence to Benefit the Consumer
Other examples can be found in the search results that Google provides, the product recommendations that Amazon offers, or the “Top Picks” for you on Netflix.
If you think back a few years ago, Google search results weren’t so accurate at retrieving results that aligned with what you were looking for, but now even if you spell the wrong word or put the words in the wrong order Google serves the content that you are looking for, most of the time. According to a 2017 article in Forbes, the first five results in Google receive 67.6% of the clicks. This accuracy is produced by the more than 3.5 Billion search results per day that Google provides and we essentially train by which results we choose, which ones we don’t, and how we refine our searches by what we search for next.
A similar process is used by Amazon to serve product recommendations and by Netflix to make entertainment recommendations.
Each of these technology companies are very good at understanding us, their customers/ users, and then watching how we interact with the results that their recommendation engines provide. By understanding things about us like our demographics (age, income, race, marital status, location, family size, education) and behaviors (search history, shopping history, timing of when we work and/ or engage online), they can then develop what are often called user segments or “personas”. These user segments and personas are used to create groups of users that can be used to predict new user’s interests and behaviors by matching similar characteristics of new users with groups of known users.
Think back to a time when you first signed up for Netflix or Amazon or a social media site like Twitter or Facebook. You were asked some demographic information and then were prompted to move through a content/ product selection process. The questions asked are ones that the tech companies “know” are the best predictors of your interests and behaviors. Your answers are further refined by allowing you to make some specific selections of products and content and voila! You log into Amazon and you are reminded that it is time to buy dog food or detergent just before your run out or to order that book that your best friend told you about yesterday or into Netflix and a new documentary is recommended about your favorite hobby.
A more personalized experience is what we as consumers not only want, but we expect and all of this effort on the part of big tech companies to understand us is paying off.
Studies are showing that 86 percent of consumers say personalization plays a significant role in their purchasing decisions. For online shoppers, 45 percent are more likely to shop on a website that makes personalized recommendations, while 56 percent of online shoppers are more likely to return to sites that offer them. (Intelliverse 2017).
So, let’s bring this all back to Augmented vs. Artificial Intelligence and to consumer-centric healthcare.
Artificial Intelligence Can Provide Consumer-Centric Healthcare and Assist in Precision Medicine
Technology companies have gotten very good at helping us find what we are searching for when it comes to many parts of our lives, searching, shopping, and entertainment. Isn’t it time that when we are searching for information for something as important as our health that it becomes easier to find what we are searching for too?
Shouldn’t we be able to take the CAPTCHA concept and apply it to diabetes or cancer or heart disease? Can’t we leverage augmented intelligence to make recommendations to physicians and patients to not replace the expertise of our doctors or run the risk of making definitive diagnoses or treatment recommendations that could end up being wrong, but instead to reduce the number of choices from hundreds to tens to inform and not replace the decision making process?
If we expect to develop highly accurate models that garner the understanding and trust that we will need to enable physicians and patients to accept them, can’t we all learn together?
This “learning together” will require that we all work together.
One great example of this learning happening in healthcare today is the All of Us research program by the National Institutes of Health. The All of Us Research Program is a historic effort to gather data from one million or more people living in the United States to accelerate research and improve health. By taking into account individual differences in lifestyle, environment, and biology, researchers will uncover paths toward delivering precision medicine. So, the NIH is using a method of learning about individual factors that impact our health as tech companies do about search and shopping. They can then create “groups”/ cohorts of people who have similar lifestyles and biology and live in the same locations to eventually be able to become much more accurate at predicting effective treatments that are more individualized and precise. They are also gathering massive amounts of what is hoped to be high-quality data.
We don’t know what all the plans are for using the data that are collected, but we could imagine that the data will be used for things like making recommendations to physicians to further investigate specific that may be contributing to a health condition thus, augmenting the physician’s decision making process. If the NIH also has access to the decisions that the physicians make, then that data can be used to refine future results training the model to become more and more accurate or for the purpose of this discussion “intelligent”.
We can stop looking at leveraging advanced analytics as an “all or nothing” experience and instead look at it as a continuum. It’s time to bring artificial intelligence and healthcare together, not to replace humans but to augment the decision-making process.
Join our mailing list to receive health innovation updates right to your inbox or Share this article on social media to engage others in the quest to “learn together” about how to most effectively and humanely use analytics to improve human productivity and enhance health decision making.