Friend nutrition -- who to see, know, ignore
/…who you hang out with will be like diet and exercise, in the future we'll manipulate body chemistry with friends.
I have a few people in my life who are a bit 'toxic'. What if I got a nudge to ignore them, or that lunch date we promised simply 'fell' off the calendar.
This is the future.
It's not about the data gathered on you, our next fixation will be the data gathered about the people around you. Human interaction feels complicated, but the intricacies of friends are ripe for machine learning. A great deal of human interaction can be patterned and captured in a very large AI system. It's coming.
I predict we will casually use the data we know about us and the people around us to manipulate our body chemistry. The surprise is that it will be easy. It will start with how people affect us. Are they toxic or a vitamin?
We already use social networks like a life-long rolodex, the obsessive of us 'curate' our feed by specifying who is a close friend, who we keep an eye on and who we unfollow. Soon, the proper dosage of friends will get served up to us in easy-to-consume social networks and digital messaging. Who we encounter may not be so random.
Why now? What has shifted in the adjacent possible to imagine this now? Growth of machine learning and neural networks, ubiquity of social media, plus the number of recent studies showing health impact of friends. Plus, it builds on what people intuitively do now.
A blast of studies using social media and real-live tracking show if you have a friend who becomes obese, you have 57% chance of tipping the BMI meter into obesity. Students with studious room-mates study more, diners sitting next to big eaters at the lunch counter eat more. Friends of friends have a surprising effect, even if you don’t personally know them, if you want to stop smoking, lose weight or survive illness WHO you know has a massive effect.
Friends, plus friends of friends. Think about the people who influence you?
Turns out the number of friends you have is actually governed by the size of your neocortex, the Dunbar number, named after the anthropologist who studied primates in the 1980’s. You basically have 150 friends. While you might have 500 acquaintances, 15 close friends and a handful of intimate friends, the 150 number is common across all cultures.
Steve Cole, a genomics researcher at UCLA found people chronically lonely have significantly more heart disease, are more vulnerable to metastatic cancer, have an increased risk of stroke, are more likely to develop Alzheimer’s, and die at twice the rate of those better connected to friends. Describing body chemistry underlying these results Cole explained “when people felt lonesome, they had significantly higher levels of norepinephrine, the hormone that shuts down viral defense but escalates production of certain white blood cells.” And he noted numerous sleep issues – so if the body chemistry doesn’t hurt you the sleep problems surely will.
Chatting about this to a writer friend, who pushed – how would this work in real life? I imagined a software application, a platform extension of a large social media network. I even went so far as to name it Friend Nutrition, with the tagline ‘who to see, know and ignore’. Offering the promise to seamlessly change our behavior – for cognitive, productivity AND health purposes.
Have you noticed how friends make you feel? Some kick you in the butt to exercise more, dress better, eat fish, think of a new project, believe a new job is possible, start a business... and others do just the opposite? Friends really are the new family, in big cities like Seattle and Denver 40% of adults live alone. New research has measured stress, inflammation, and sleep in response to your friend connections... all the things that affect how your body fights disease. Experience how the underpinnings of your immune system could be bolstered by the people around you.
This imagined app gives users an invisible nudge – who to see, who to know and who to ignore. The algorithm ensures which friend’s posts have visibility in your feed, which text messages always get a response, which alerts take priority, it watches your calendar and makes sure to rearrange for people you need to make time for. Those lunch meetings with toxic people get rescheduled enough times to fall off the calendar for good. And network mathematics influence which people you need to meet and pushes you to get to that event. The feeling of random encounters for that perfect introduction at the right moment is not so random anymore. Amidst a day of too much to do and too few real friend encounters - Friend Nutrition is an invisible nudge that pays off.
The potential pitch to venture investors is - software to disrupt the random way we make friends by using machine learning and voice sensors in your phone. People want to connect - 81% of US internet users are on facebook, 50% of 18-24 year old’s go to social media when they wake up everyday. Friends are the most powerful mechanism to influence how you feel – and it’s left to chance? Friend Nutrition will watch, learn and serve a nutritious mix. No FDA approval required.
If there is an inch to be found in anything that is random, disorganized or could be tweaked for advantage, the software world will find it.
The most ubiquitous sensor voluntarily carried around is the microphone in your phone. Voice pitch, timing and natural language analysis can indicate emotional response. Think I'm making this up, check out the Sociometer study from MIT, now called Humanzye, captures body mirroring and voice pitch to good result. The language analysis tool LIWC analyzes psychographics of the speaker and accurately models two person interactions. Another AI project at MIT Media Lab monitors emotional tone of unstructured conversations in real time with 83% accuracy, and a sentiment score on 5 second chunks of dialogue. While there are tons of sensors possible, voice and text alone could be enough to monitor emotional response of people interaction with high precision.
How can you measure an actual friendship, that’s nuts. The AI in this app doesn't actually understand HOW you make friends, the algorithm just watches your response to people. Assuming you’re limited to about 150 relationships, friends are often one dimensional - we don’t insist they be everything like a parent or a spouse. One friend invites you to fun parties, another challenges you to be a better person. Does a finite number of one-dimensional friends make it easier to model and adjust a mix?
The AI in this app doesn't actually understand HOW you make friends, it just watches your response to people.
The learning algorithm can observe steps that lead to a positive friendship and favor more of those. The right mix is more like a ‘cocktail’ sweet, salty, acidic… Not all our encounters need to be super-pleasant, A perfect formula could be a mix of slightly acidic questioning people who make us nervous and others who act like a positive vitamin boost. And every now and then you need to meet an "I-learned-a-lesson" person, or you’re doomed to be one of those naïve unchallenged souls.
This whole system of an unsupervised machine learning for friends would be like AlphaGo from Google. It mastered the complex game of Go. The intricate system of moves, countermoves, and how a friendship forms, shifts and responds is something like a game. AlphaGo for friends. Watching latent variables, the analytics could find unnoticed or uncharacterized human responses to friends in the real world. The training dataset for billions of social interactions exists, WE have been training the social media dataset for a decade - it's huge, 510,000 comments a minute. 2.2 billion facebook users. Facebook is learning how we interact, better than we understand ourselves.
Analytics of friends - how could it benefit us in a way that’s helpful enough to overcome the slightly invasive manipulative creepiness of it. You can see it would work, a deep learning algorithm finding a way to payoff what sociologists have been touting, layering on vast social media use and reliance on digital tools for most all connections. It’s a drug-free boost using our own human chemistry.
Perhaps this scenario is not so far from reality, our friends will become like diet and exercise. I wrote this article because I imagine an academic team somewhere is working on this right now and could use some encouragement. Tell them to feel free to steal the name and tagline.
