And if, that’s not possible, you can at least exercise some control because you’re tackling a, heterogeneous collection of short tasks, you can also employ directly assess whatever is. about which games, you choose to play. Then he's adding benefit b to every situation where you take holiday. He is the author, with Tom Griffiths, of Algorithms to Live By, a #1 Audible bestseller, Amazon best science book of the year and MIT Technology Review best book of the year. I would love to know about efficiently but roughly sorting material. The chapter provides some evidence that humans tend to over-explore. If that holds, and if you are limiting the amount of time to get a workable model, you should be able to constrain yourself to simpler models. all in one go when the, Consider how many times you’ve seen either a crashed plane We can look at algorithms as case studies in rationality. And, indeed, people are almost always confronting what computer It may just be that I'm at the tail of the distribution on explore vs exploit. At first it seems like learning a new skill would be another example: you can learn different subskills or use different instructors. the rarity of those, lags is a testament to how well you’ve arranged it: keeping Discussion in this chapter has pushed me closer towards regularly timeboxing. I guess that makes sense. For example, you’ll “explore” the area you’re in while you have time, trying new local places. From finding a spouse to finding a parking spot, from organizing one’s inbox to peering into the future, Algorithms to Live By transforms the wisdom of computer science into strategies for human living. If you expect to search through a large number of items a small amount of times, then you may not be much better off having sorted than just searching through it. Before moving to a new location, however, you’ll “exploit” the results of your exploration by revisiting your favorite places. Like really large. The most prevalent critique of modern communications is that The explore/exploit tradeoff tells us how to find the balance between trying new things and enjoying our favorites. You miss an episode of, your favorite series and watch it an hour, a day, a decade Or one with a high expected value? You can download Algorithms to Live By: The Computer Science of Human Decisions in pdf format The baseline is taking no holiday in a low holiday environment. We need solutions that trade off integrating knowledge of the tree, future options and the cost of spending time thinking. In English, the words “explore” and “exploit” come loaded with completely opposite connotations. Probably, this is a good thing. When n is 1, the Erlang distribution collapses to the exponential. When you cook a bread from a recipe, when you knit a sweater from a pattern, when you put a sharp edge on a piece of flint by executing a precise sequence of strikes with the end of an antler- a key step in making fine stone tools, you are following an algorithm. Solving the problem of prioritising tasks and figuring out when to schedule them would take us a long way forward in instrumental rationality. Brian Christian is the author of The Most Human Human, a Wall Street Journal bestseller. Boris Berezovsky. What constant you ask? disproportionate, occasional lags in information retrieval are a reminder of it’s very hard to predict the future. I normally think of timeboxing as a method for breaking down a task and reducing the delay until payoff. The literature on over-exploration is the strongest reason to think I might be wrong here, but there's also a threat from something like social desirability bias. I hadn't thought of this as a way to generate simplicity before. which, if true, seems reasonable evidence of its suitability for adoption. As I can program, I intend to look into making tools for myself in this space. This is not merely an intuitively satisfying compromise Well, if b + h > s, or if b - k > 0 then not taking holiday no longer dominates. gunfire—the amount, of confrontation quickly spirals out of control as society Rather, for values much less than the mean, it's safe to assume the mean (or just over). If we're thinking of a reading or a todo list, a human would rarely work through it in order, but would keep an eye out for high priority items (a counter-example for me is RSS: I often do churn through my feeds in order). When, our expectations are uncertain and the data are noisy, the This chapter discussed its role in keeping work limited when marginal payoff becomes uncertain. If b + h > s, but b - k < 0, there are now 2 equilibria. Including hiring, dating, real estate, sorting, and even doing laundry. The Erlang distribution generalises this to the time it takes for n such occurrences. But I hadn't drawn out the specific implication from low number of interruptions to vanishing hours. If we model that as a constant addition to the logarithm, (as in log(expected) = log(observed) + log(k) = log(k * observed)), then we recover a multiplication heuristic! If you haven't, first think of the exponential distribution. 1. This could help a lot with explicit estimates and making predictions. They won't help you update your belief about the mean of a normal distribution, nor that it looks more like an Erlang distribution than a power distribution. But I still think that allowing long lists in my life is a problem. Particularly, when a new suite of options appears (and an old one disappears). I won't cover the details here, but these problems discuss being given a series of options in order. What about if the CEO pays people take holiday? If b - k > 0, but b + h < s, then there is no longer any equilibrium at all! Even in quite transferable cases, like sorting, it pays to remember a piece of old programming wisdom: Rule 3. This is payoff h. As discussed, not taking holiday dominates taking holiday if s > h. This leads to a bad equilibrium: one where no one takes any holiday. This comes from this chapter claiming a cache-management algorithm called LRU (Least-Recently Used) performs well in a variety of environments. exhaustively, enumerating our options, weighing each one carefully, and We model the rest of the company as a single agent taking a 'high' or 'low' holiday strategy. Explore vs. what would you do if you could not fail? There are two problems with leaving more things unsorted: You might not have a good intuition about which things you look through often and which rarely. Nonetheless, I found it a useful lens to think with. But as soon as everyone is, it pays to defect! That is, exploration is considered laudable and cool, so I will be drawn to increase my exploration to also increase my status. small-scale groups; they, do in nature. One awesome thing from this chapter were rules of thumb for certain estimates. I think this is an improvement but I'm not that confident (maybe around 3:2 that it's an improvement). Algorithms to Live By: The Computer Science of Human Decisions by Brian Christian and Tom Griffiths There are predictably a number of readers who will look at this title and shy away, thinking that a book with "algorithms" in its title must be just for techies and computer scientists. just how much we. But I've found the feeling that is has made concrete and explicit some of my intuitions about work valuable (for example spending a lot of time organising reference materials seems to always be busy work that doesn't pay for itself), It accepts and explores differences between our decision contexts and those contexts for which the algorithms were developed. For this to work, you have to actually explore simpler options first, which one might not lean towards instinctively. science regards as, the hard cases. The rest of this section is a concrete expansion of the reasoning behind computational kindness. In Algorithms to Live By, authors Brian Christian and Tom Griffiths devote an entire chapter to how computer algorithms deal with the explore/exploit conundrum and how you can apply those lessons to the same tension in your life. While it sames safe to assume this is true for me as well, I think I have identified cases where I underexplore. But if you force yourself to actually come up with a model/solution in the time allotted, you are very likely to lean on simplicity. simplification by stroke size: Unless we’re willing to spend eons striving for perfection Algorithms to Live By. But in practice, when the clock—or the ticker—is To be concrete, one way to control how complex your models or plans are is to restrict the amount of time you allow yourself to generate them. Here are the three changes I've made that have been most worthwhile so far: When I first get a set of new options that is likely to stay stable into the future, I prioritise choosing a new option over repeating a good choice (from Explore / Exploit). connected; we’re. These rules have the potential to be useful, but they don't give you much guidance for shifting the distribution that you think underlies the phenomenon. people in the United, States killed in car accidents over that same time is Overfitting is what happens when your model is too sensitive to idiosyncratic details of your training data. But note, that's your expectation of the total amount! Or, when writing a lengthy post, you could indicate at the beginning where to find things that might be of interest to the reader (obviously I've not done this perfectly, but I hope the intention is well-taken). You can either play a strategy of taking holiday or not. (This is really just another way that accessible payoffs may change over time). The feeling that one needs to look at everything on the It’s this, that forces us to decide based on possibilities we’ve not I imagine I'm not alone in the face-reddening experience of scrabbling through pages of notebooks and folders full of loose-leaf documents in meetings while everyone looks on. If you unilaterally take holiday here, it turns out badly for you. difference is enormous. Gwern has produced some practical prior art. A leap from ordinal to Sorting & searching are perhaps the most archetypical algorithmic activities, and these chapters did a fairly good job of expressing how much approaches could differ in efficiency. So we can apply the rule for the normal distribution: if the logarithm of your observation is significantly less than the logarithm of the distribution's median (so let's say the observation is about half the median) just go with the median. I've gone a long time without zeroing-out my reading list. best bet is to paint with, a broad brush, to think in broad strokes. Or am I missing a point here? A fascinating exploration of how insights from computer algorithms can be applied to our everyday lives, helping to solve common decision-making problems and illuminate the workings of the human mind. I would consider it evidence against the book if it claimed it had lots of high value, very novel advice. It just has to be unambiguously better than your competitors, even if all of you could have done work that was 90% as good in 10% of the time. That is, add a fairly small amount relative to the scale of the distribution. equilibria, and information cascades. I’m not sure what I can take away from these algorithms and apply them in my daily life but this was a fun read for me. Keeping gym items in a crate by the front door. In contrast, the number of Optimal Stopping — When to Stop Looking; Explore/Exploit — The Latest vs. the Greatest; Sorting — Making Order The In almost every domain we’ve considered, we have seen how Third, Exploit. But that seems to me to be correct for the specific case where the Erlang distribution reduces to an exponential distribution, and not the general case. Algorithms to Live By takes you on a journey of eleven ideas from computer science, that we, knowingly or not, use in our lives every day. We may get similar choices again, but never that exact one. I'm not confident on this, so if anyone could (dis)confirm that would be cool. then selecting the, best. This chapter discussed some algorithmic approaches to that problem. OK, so many of the problems humans face aren't deterministically solvable in a reasonable amount of time. If you don't have long, stick to exploiting; if you have years, shop around. By being concrete and proposing specific actions or times, we can allow someone to only check rather than search. the more realworld factors we include—whether it’s having incomplete Algorithms to live by: Explore vs Exploit “Trying new things or sticking with our favorite ones?” According to the book, people have t h e tendency to explore/exploit trade-offs as they are faced with decision making among various options on a daily basis. As humans, as well, we can be prone to adding an extra detail to our model: a complication we think we should probably account for. latencies, take heart: the length of a delay is partly an indicator of the extent intractable recursions, bad. But that’s almost never the case. TL;DR: check out if you should explore something new, or exploit a favorite! But piecemeal accounting for complications is dangerous. Increasing the cash on the table in the prisoner's dilemma, for instance misses the point: the change doesn't do anything to alter the bad equilibrium. home; you’re just, calibrating. That is, there are values at very different scales than the median. Little evidence is provided (an article by a Dropbox intern and an early paper on caching) beyond the claim that LRU is great, so I would have to do research to back this up. Tversky & Warren, 1966) studied situations that particularly favoured exploitation. If you If you are not familiar with the problems of interruptions, Cal Newport's written a lot in this area. Table of Contents. Because of a discussion of an idea called 'buffer bloat', I became keener to reduce the number of items on my todo & reading lists. optimal stopping problem is the implicit premise of what it reference to a common quantity. But without exploring, there's nothing to exploit. You don’t know the odds in advance. without, decreasing your responsiveness below the minimum acceptable or a crashed, car. I'm not certain whether that should seriously reduce my confidence or not though (the hypothesis still has be relying on advice I evaluated well before). Explore vs exploit honesty is the dominant strategy. when trying to, resolve the explore/exploit dilemma, or having certain tasks If this book offered large undiscovered gains in this area, I'd be pretty uneasy. In a few paragraphs there's a reader's guide so you can skip around. Sorting theory tells us how (and Odds above 9:1 / 90% confidence that this has been an improvement, but I have doubts about its long term feasibility. Yes, a lot of its advice is already encoded in my intuitions or in folk advice. There is a tension between getting value from the best known option ('exploiting') and checking to see if there are still better options ('exploring'). I just can't do the weekend and the week after next is less good.". Contains mathematical philosophy on decision making on a wide range of topics. individuals sharing the same. But as you gain more knowledge, you lose some opportunities: branches get left behind as you follow the track of waiting and thinking. Though the book is flawed, I have changed my behaviour in some ways because of it, and am considering changing others as well. The Secretary Problem. limit. In networks, this can lead to the receiver thinking the sender takes a long time to receive and process responses. For example, the authors discuss the game-theoretic problems with unlimited vacation: assuming you get some important benefit by taking little holiday relative to everyone else, and pay some cost by taking the most holiday, the equilibrium here is with no days of holiday (assuming the costs and benefits are large compared to taking a holiday). I'll get to them in a second. were probably on, another continent, transmitted to you via the Internet or If we imagine that everything is falling into the equilibrium in this scenario and everyone has the same payoff matrix, we can just imagine this as everyone takes holiday or not in unison. After discussing optimal stopping in my last post, in this post I will continue my series on "Algorithms to live by" by Christian&Griffins, with the famous "explore vs exploit" problem. If the logarithm of the observation is significantly greater than the median, we expect the logarithm of the final elapsed duration (or what-have-you) to be a bit bigger than the current logarithm. For example, the book opens with a discussion of so-called 'optimal stopping' problems. grows. correctly, this is not, just wishful thinking, not fantasy or idle daydreaming. What about if we ask for an OK solution? It is possible to be extremely astute about how we manage difficult decisions. The classic comparison between bubble sort and merge sort really pumps up your intuition that there could be hacks to be found! the appeal of, lesser-known options beyond what we actually expect, since Keeping hoover bags behind the sofa. And if b + h > s and b - k > 0, then taking holiday becomes the dominant option! Shifting the bulk of one’s, attention to one’s favorite things should increase quality Sharing points: 1. In general, however, it seems I should be increasing my tendency to exploit. In the book Algorithms To Live By, Christian and Griffiths show how much we can learn from Computer Algorithms.The book goes over many algorithms like Optimal Stopping, Explore/Exploit, Caching, Scheduling, Predicting, Networking etc. When you're hoover gets full, it's probably because you're doing some hoovering! I am more sceptical that this generalises. Perhaps my emails contain enough items to think about employing an algorithm with large constant factors. In general, I think better introductions are available in the LW-o-sphere, for example this recently curated post. cardinal. between looking and leaping. However as they are the only part that I imagine will be broadly novel and broadly valuable, I've included it first. Everyday low prices and free delivery on eligible orders. Linearithmic numbers of fights might work fine for One idea the authors cover seemed particularly useful to me: early stopping. Starting from every moment, there are choices you could make. benefit the rest of the time by having what we need at the Imagine walking into a casino full of different slot machines, each one with its own odds of a payoff. time period the presence of gun violence on American news When you’re truly in the dark, the best-laid plans will be memory, Ramscar says, should help people come to terms with the spinning our tires we, imagine easier versions and tackle those first. Algorithms to Live By by Brian Christian & Tom Griffiths is an exploration of the applicability of algorithms from computer science to human decision problems. If you take holiday in a low holiday environment, it costs you k. It could be the case that k = s, or even that k < s, though we probably imagine in most cases that k > s. If you take holiday in a high holiday environment, you just get to enjoy the holiday! pleasant surprises, can pay off many times over. Nonetheless, the figure seems reasonable enough that I feel comfortable using it as a motivational bump. If you're lucky, it will tend to happen in the same place as well. Algorithms to Live By by Brian Christian & Tom Griffiths is an exploration of the applicability of algorithms from computer science to human decision problems. You understand the company better if you have worked with multiple teams. I mentioned the book has flaws. Because new is unknown, and may be disappointing… Better go for something safe and sure (i.e., exploit). The explore/exploit tradeoff tells us how to find the balance between trying new things and enjoying our favorites. To see this, remember that the logarithm of a lognormally distributed variable is distributed normally (hence the name). As he puts it, “A lot, of what is currently called decline is simply learning.”, Caching gives us the language to understand what’s Think long and With overfitting, you end up predicting that data will at each point err from the 'true average' in the same way that the data you sampled did. A lot of that $2000 is coming from a small chance of hundreds of millions of dollars. Tom Griffiths is a professor of psychology and cognitive science at UC Berkeley, If you want the best odds of getting the best apartment, The median should be less than that. Second, from the Metropolis Algorithm: your likelihood of following and all—to the very first place you see that beats whatever prediction rule is, appropriate—you need to protect your priors. Here's a blog post of his that came up when googling "Cal Newport interruptions". The explore/exploit tradeoff tells us how to find the balance between trying new things and enjoying our favorites. out of a totally, random state, using ever less and less randomness as time Consider only reading the introduction. And to end up in a, situation where finding the perfect solution takes One at everyone taking holiday and one at no one taking holiday. Merrill Flood. A problem with this section (of this post, as opposed to of the book), is that I don't feel like I had many small insights I can summarise. It’s [...] The problem isn't that vacations aren't attractive; the problem is that everyone wants to take slightly less vacation than their peers, producing a game whose only equilibrium is no vacation at all. The problem is information getting stuck at the back of a long queue, with the sender none the wiser. space, requires a leap beyond. Let’s start with a working definition. While this isn't the most satisfying rule, I could see it providing some use in Fermi-style estimates and hopefully my intuitions about it will sharpen. Book Summary – Algorithms To Live By :The Computer Science of Human Decisions. greater than the entire, Simply put, the representation of events in the media does You can pass forever on an option or accept it and see no more options. If it had been new to me, it would have been quite valuable (though probably still not enough to move this section up). Whether it's revisiting a course of action that seems worthwhile but more-and-more likely to fail, checking to see if a software build is done, or attempting to schedule dinner with a friend that's always busy, simply doubling the interval between attempts seems a reasonable first stab at keeping information-gathering costs down without giving up on promising avenues. all human knowledge is uncertain, inexact, partial. television. There might not be much better to do while sorting and there might be large advantages to being able to find the relevant material quickly. unreasonably long. In some situations, spending more time in total sorting and searching is a good choice. Getting from the bad equilibrium to the good one is ... difficult. It may well be better to pay someone to do this for you. It could be seen as failing to prioritise simplicity in your models over ad-hoc additions to capture exceptions. “I think the most important tangible thing seniors can do is to try to, get a handle on the idea that their minds are natural metrics might be just, as important. For completeness, I will give some concrete sorting algorithm suggestions. Or one that's probably good? the costs of error, against the costs of delay, and take chances, Book Summary: Never Split The Difference Summary By Chris VossBook Summary: When Daniel Pink SummaryBook Summary: Rejection Free Summary Scott AllanBook Summary: The Universal Law Of Success Summary Albert LaszloBook Summary: Unfuck Yourself Summary Gary John BishopBook Summary: How To Stop Feeling Like Shit Summary Andrea OwenBook Summary: How to Fail at Almost Everything Summary By Scott Adams, No time to the whole book ? I've not taken action on this yet.