What are the dimensions that people care about in project selection

# What are the dimensions that people care about in project selection People care about many different dimensions of a project. Awkwardly, each of these is dependent on internal preferences, so everybody’s evaluation is going to be different. Additionally, even if those preferences can be stated concretely, like “expected number of good increased human life-years” it is subject both to definitional nebulosity and outcome uncertainty. Unmeasurable metrics that depend on non-consensus discretion create definitional nebulosity — what entails a good human life year? How much more or less good would a hedonistic VR existence be compared to one spent on an FTL ship to the Orion Nebula? These questions can only be answered based on individual preferences. Processes with not only unknown or unknowable outcome distributions but unknown/knowable *dimensions* to their outcome distributions have outcome uncertainty — arguably the most significant legacy of the atomic bomb and space race was getting the American semiconductor industry through the valley of death. Note that we can’t even tease apart causality *in retrospect*. Another important piece of the calculus is the fact that [[Most predictions suck]]. So timescales and effects of any future-focused project are going to at some level be based on human intuition. Without preferences and intuition everybody’s work and all money should go towards enabling infinite upside scenarios or preventing infinite downside scenarios. If you don’t think there are infinite upside or downside scenarios, you just used your intuition or preference. If you accept the fact that preferences and intuition are unavoidable, any system for looking at potential project prioritization needs to embrace the fact that it needs to account for individual preferences and human intuition is going to slip into the system eventually. The question is where and how do those preferences enter the system. I would argue that there are better or worse places for intuition and preferences to enter the system. I suspect that it is better for intuition and preferences to enter the system as close to the final result as possible because that prevents them from being obfuscated. Following this principle, I suggest that any final score be a weighted combination of taste based scores and intuition-based scores that area all explicitly come from people. A bookkeeping aside: Areas should be separated from their goals which in turn should be separated from projects. If areas are not separated from their goals/progress metrics the system will favor more definite areas which should be an explicit preference decision, not a hidden one. ### Area Based Evaluations Taste-based scores. These are subjective numbers that are actually at the core of people’s preferences. Maximizing these in the world tend to be *intrinsic* goals. * Wonder - how much wonder would this introduce to the world (as a first order effect?) Being able to see in person a moon rise over a ring of Saturn or discover a new law of the universe seem to maximize wonder while living longer (more of the same) would seem to minimize wonder. * Definiteness - how clear are the goals of this area? How easy is it to know whether you’re making progress? Increasing average human lifespan is extremely definite - it can be measured directly. Atomically precise manufacturing is not very definite - we have a sense that we want to be able to put each atom exactly where we want, but what that entails and what it would allow us to do and the path to it are all extremely fuzzy, even from Drexler. * Abstraction - how removed is this area from intrinsic goals (which of course are subject to preferences, but can be abstracted a bit)? Increasing human lifespan is very close to intrinsic goals - most people do not want to die. Filling the niche in the innovation ecosystem that corporate labs once filled is much more of an instrumental goal so it would have high abstraction. * Quality of life - how much would progress in this area directly affect quality of life. Health is huge here, as are areas that directly reduce suffering. An area that makes things that increase quality of life (like energy) cheaper are probably in the middle. Note that this is *separate* from wonder. An early Marian colonist would have a low quality high wonder life. * Human Capability - how much does this area increase our collective ability to manipulate the universe around us? Engineering-focused areas like APM would rank very high here and human health would rank low. * Equality - how much does this disproportionately help people on the lower end of the socioeconomic spectrum? Better desalination would rank highly here, accelerating science would be low. (Historically, many unequal things have equalizing effects but this is just about predicted first-order effects.) Intuition-based scores. These are objective numbers that are based on someone’s intuition for an underlying fermi problem. Maximizing these in the world tend to be *instrumental* goals. * How many years could we expect until this has an effect on the majority of people’s day-to-day lives? * How much quantifiable value could it create? * What order of magnitude would it cost to unblock this area? (This one is especially important because arguably while general AI is infinitely impactful, it would also be infinitely expensive to * ### Project-Based Evaluation Intuition-based Scores * How much would this cost? * If this project were wildly successful how much would it go towards unblocking its area? (0-100%) * What timescale would you expect to see an outcome on? * How directly does the project address the areas goal? A crazy way to do it would be to just use a machine learning model to learn someone’s preferences by present a whole bunch of projects and having someone rank each one on a scale of 1-10. You then turn a written description of the project into a feature vector and train a custom preference model for each person. This is not a good idea.  [Web URL for this note](http://notes.benjaminreinhardt.com/What+are+the+dimensions+that+people+care+about+in+project+selection) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/What+are+the+dimensions+that+people+care+about+in+project+selection)