Originally posted on Medium.
For a high-level discussion on this topic, check out the NonTrivial Podcast.
Contribution as Survival
Life doesn’t just happen. We must carve out an existence by taking actions that facilitate our survival. We make decisions each day that hopefully move us in a fruitful direction. We interact with others, take jobs, produce value and get compensated. There is a cost to becoming fulfilled, and that is the time and effort we put into our contribution to society.
A job is largely how we make that contribution. Our economy puts in place a framework for the division of labor, so that people can add to the things companies build. The economy rewards us for our time and effort by way of salary or business revenue. Through the years we acquire more skills, increase our earning power and deliver greater levels of value over time.
This contribution is what the world expects from us. By fitting into the system and playing our role we help society function. Doing something for the greater good brings us benefits related to quality of life. We get to drive on solid infrastructure, consume safe foods and walks the street feeling secure. In short, our survival depends on our ability to contribute to the economy.
And just as society expects things from us, we expect things from society. Contributing to the economy grants us the freedom to pursue our own interests. Making money means we can raise families, go on vacations and enjoy drinks with friends. Time and effort are the currency of freedom.
Whatever it is we’re searching for in life, we depend on society to get there. If we fail to meet what society expects from us there is little reason to believe we will find contentment. Finding meaning in life is not a solo journey; it requires we participate.
The Individual and the Group
Most of us accept that our contribution to the economy is what grants us the freedom to pursue personal interests. But working-hours take up most of our day. The real goal in life is to find employment that is itself fulfilling. This is why we find ourselves trying to get the job or education that both aligns with our interests and allows us to make a good living.
But in reality this is difficult. While the job or degree title may reflect our passion, the day-to-day is often anything but. The expectations of society cannot be tailored to the individual. Society is a group phenomenon. Expectations are created out of averages, no different than any institutional approach to standardizing the inputs of its members (e.g. scholastic achievement).
The mismatch between individual and societal expectations is necessary because specific details always subsume into higher level aggregates that look nothing like their parts. An individual is a specific instantiation of something more general. There is no reason to expect that which exists in aggregate should look like that which makes up that aggregate. This is why individual needs don’t truly map to aggregate level social structures. People are individualistic and nuanced. Their interests follow idiosyncratic reasoning, not socially aware ideals.
Despite the individual-to-group mismatch we need to find a way to strike a balance. We must conform to the social constructs that make our lives possible, but we also need to feel valuable, and live lives that are meaningful. Any happy life will exist at an intersection between what we expect from our world and what the world expects from us.
Solutions at the Intersection
Most of us would accept that life is about balance. It is rare that we can take an absolutist approach without consequences. We need to make concessions in life to be both true to ourselves and accommodating. Life is nuanced; neither black nor white.
Think about about any human interaction that looks to make progress by bringing together ideas. Most of us are unlikely to agree on the fine points, but at some level of abstraction we must recognize a mutual set of values. A completely dualistic take cannot move us forward as a group.
A common setting where this plays out is when we bring our opinions to the table at a company meeting. It is unlikely everyone in the room is going to agree, but it’s important that some of our ideas are taken into consideration. Working effectively in society means existing at that intersection between wants and expectations; the big question being, “how can we do this effectively?”
The first step is to realize that structures exist for a reason. There are no lack of things to rightfully complain about in our education, legal and health systems. But in the complete absence of structure society wouldn’t function. Structure provides a place for our ideas and creativity to land on. The best parts of those structures are the ones that have lasted centuries if not millennia, as timeless truths.
And so whatever solution exists at the interface between expectations requires us to bring ourselves to existing structures. We need to find ways to deliver on what’s expected from us without compromising who we are.
Finding the Joint Distribution
Whatever is being expected of us can be thought of as a set of outputs we’re expected to deliver. It might be the maintenance of a piece of equipment, a quality scan of a patient’s body or a presentation on strategy to stakeholders. Whatever our vocation there are people on the other end of our actions expecting our outputs.
The problem arises when we feel it’s unnatural to produce these expected outputs. While a good number of them should be aligned with our interests (why else are we taking this job) there are always ones that do not. Say you take a new management position in a software company, after spending 6 years writing code. Maybe you took the position to get a more holistic view of the company’s work and its impact. Maybe you’re good at thinking strategically and believe a managerial position would allow you to flex this muscle.
But as you settle into your new role there are expectations you find mundane. While you still get to be strategic, there is also setting up and attending countless meetings, giving boring presentations, putting out fires and regularly interviewing candidates to help the company find talent. In any job or career there are many expected outputs that don’t align with our interests, yet failing to deliver them would be detrimental.
Just “sucking it up” never works in the long run. There is nothing more soul-sapping than showing up every day to a job you can’t stand. It takes us away from ourselves. Getting paid is never enough; we need to feel fulfilled.
In order to not compromise ourselves yet still deliver outputs misaligned to our interests requires that we dig deeper than the superficial labels attached to expectations. Human activities are not singular things. They are better thought of as distributions of actions. That new management position had a set of mundane objectives, but those are just a few of the activities expected in management. If we deconstruct management into a distribution of possible activities it might look something like this:
There is no definitive definition of management in the real world; as with all things, meaning requires context. It is better to think of management as a label that is arrived at by a distribution of various activities.
Since we are talking distributions, and thus thinking in terms of probability, we can think of observing someone in management, and seeing various actions (e.g. scheduling meetings, delegating work, etc.), each of which have a corresponding probability of occurrence. All the actions taken together (the “sample space” …. as in all possible outcomes of an experiment) would constitute the meaning of management. Once you get into the nuances of a specific company’s culture the activities associated with their version of “management” is for all intents and purposes a random variable, and we should expect the sum of all the possible probabilities to be 1:
This isn’t to highlight some obvious tenant of probability but rather to encourage the looking upon labels as coming from a distribution of possible activities.
We can take the same approach with the area we are passionate about. Perhaps your 6 years in software development reflected a genuine passion for building things. Regardless of your reasons for wanting to now view things holistically, it’s building things that truly drives you. Let’s be a little more specific, and say it’s building AI “data products” that gets you up in the morning. Deconstructing “data product development” into a distribution of technical activities might look like this:
The thing about probability distributions is we can add them, multiply them and find differences between them. This mixing and matching of distributions reveals something about the inherent uncertainty of situations, and uncovers information that can be used to make decisions.
We can imagine many ways we might operate on our management and development distributions above. Relative/conditional/cross entropies, mutual information, Bayesian updating, available work, various divergences and even topological spaces all have something to say about the probabilities of occurrence that might exist at the intersections of these distributions.
Of course actual calculations would require some knowledge about the type (shape) of the distributions we are interested in, but that’s not what’s important here. There is something critical we can say about distributions in general when they get combined; something that can help us find the connection between different domains.
Any 2 distributions could possibly form a joint probability distribution. A joint probability distribution occurs when we can form a probability distribution on all possible pairs of outcomes. Figure 3 shows the idea of a joint probability distribution between our management and development distributions:
The new distribution formed in the center represents the possible pairs of outcomes. In other words, there is a chance we would observe both hiring candidates and creating a user interface. As with any distribution we can expect the sum of all the possible (combined) probabilities to sum to 1:
But what would it mean to observe both the hiring of candidates and creating a user interface? Consider the following situation. We take the new management job and one of our expected outputs is to work with human resources to help hire new talent into the company. This requires pouring over submitted resumes, analyzing them for relevant content and justifying the eventual hiring of the best candidates. There’s a good chance none of these sound particularly interesting to you. But what if it were possible to achieve these outputs as a byproduct of doing something we are more interested in? This would require that both the hiring duties I mentioned and the activities we find enjoyable co-occur.
They might both occur if one activity was achieved as a byproduct of the other activity. I am not talking about “transferable skills”, which are skills that are portable across different jobs, like being good at communication. Here I am talking about being able to achieve the expected output by working on something entirely different.
By working on the creation of a user interface we can achieve pouring over submitted resumes, analyzing them for relevant content and justifying the eventual hiring of the best candidates. We can do this by creating an interface that allows one to upload resumes, have the text extracted and analyzed for relevant keywords, which get visualized into charts and graphs to communicate why someone should or should not be hired.
The kind of co-occurence is all too common, although we might not think of it in terms of mixing distributions. Think of learning any new skill effectively, like speaking a new language. This works best not by learning individual words but by immersing ourselves into the actual culture. It’s when we are having fun that we end up learning things as a byproduct.
In reality there is a finite probability of 2 observations occurring at the same time when we have fun, because there are always ways to make actions we don’t care about be a part of ones we do.
I don’t find baseball particularly interesting, but I have friends who do. There is a kind of expectation that if we get together we speak about the names and statistics of players. I could sit and nod my head to feign interest, or I could find a way to meet these expectations by doing things I find more enjoyable. Perhaps it’s those technical activities shown in Figure 2 that I truly enjoy. Is there an joint distribution between those activities and baseball?
Of course. Tasks like selecting and preparing data in the course of training an AI model would co-occur with player names and statistics if those were the data that fed my model. I can imagine creating a baseball application that allows users to input their favorite players and see AI predictions on their chances of winning, maybe visualizing their stats.
We don’t have to change who we are to achieve outputs we are not interested in. Those outputs can be arrived at as byproducts to a target more inline with our interests. It doesn’t matter which 2 domains we are comparing; we should expect there to exist some joint distribution between them, which provides the kind of “byproduct solution” that can make you effective at personally misaligned tasks.
Delivering What’s Expected, Our Way
I argue that learning things as a byproduct is how genuine learning happens all the time. Both my friends and me learn baseball names and statistics as a byproduct. My friends enjoy watching the game, and only learn the names and stats via natural exposure. I enjoy creating models and learn the names and stats by wrestling to make models work. In both cases our knowledge emerges as a byproduct of something we are actually interested in.
One of the deepest problems with the education systems is how it tries to teach others directly, rather than having students learn through context. As I discussed in my last article, math education is the perfect example, where sterile symbols stripped of context are presented to students who couldn’t care less. Teaching someone the stripped-down version of knowledge leads to shallow learning at best.
Learning comes from building things. We have to bring something into the world. Rather than merely thinking about how your skills transfer to some other domain, think about what you love to create, and how internal activities might co-occur with a set of socially expected outputs.
We tend to make a distinction between “fun” and “work” but this is a mistake. Things are fun for a reason. What we find fun is an emotional response to activities that drive us, that make us want to learn. When things are fun we are the most effective we can possibly be. There is no reason not to have fun.
Whatever expectations society places on us, we can meet them our way. Find the joint distribution where outputs co-occur, and you’ll find ways to contribute to society without compromising who you are.