This is a linkpost for neelnanda.io/blog/48-rating

I often get paralysed by indecision when I have a bunch of tasks to do - it’s hard to figure out where to start! My standard solution is sorting the tasks into priority order - usually implicitly with a comparison-based sorting method where I go through the list one-by-one and put each new item in the right position. But I’ve recently gotten a surprisingly large improvement by rating - giving each item a score out of 10 for priority, and going in order by score. I have no idea how much this generalises to other people, but I’m sharing this quick post making the case for why this method is better in case anyone else vibes! (note - this post contains major amounts of overly opinionated overthinking of simple ideas! Read on at your own risk)

Reason 1: I avoid thinking about the hardest comparisons

Unlike in standard computer-implemented sorting algorithms, comparing any pair of tasks does not take the same amount of time. Most comparisons are obvious, while comparing close tasks feels much harder and more ambiguous. In practice, putting a task in exactly the right place among similar tasks is what absorbs the most time and attention when priority ordering, and is where I most easily get discouraged.

Rating, in comparison, feels easy for any task - I can just convert my gut feel into a number in a few seconds. If it feels ambiguous, eg I can’t decide between a 7 or 8, I just put a 7.5! Ratings can be continuous. And then, if I end up with several tasks with the same rating, I can just arbitrarily tie break (eg, by the order that they were entered into the list) - if they have the same rating, then prioritising further can’t be that useful! In hindsight, the hardest comparisons are in fact the least useful - if I do the fourth most important task before the third most important, that’s no big deal, but doing the tenth before the first is a screw-up!

Reason 2: Rating means going through the list once, sorting goes back-and-forth.

I find that this makes things much faster, and cognitively easier - to sort each task into the list I need to read back through for each item to find the right position, which is distracting and flow-breaking. While for rating, I just read through once and consider each task in isolation

Technical aside: I’m comparing insertion sort, which is quadratic time, with bucket sort, which is linear time with a worse constant, but I think time complexity is a distraction here. I rarely priority order lists with >20 elements. Fancier O(nlog(n)) sorting algorithms like merge or quick sort might be better, but seem too high overhead to be worth worrying about

Examples: I find that I need to priority order tasks lot, often multiple times a day, so getting better at this is valuable to me! Eg ordering tasks at the start of each day, processing my to-do list, deciding which of my bugs to focus on in a coaching session, deciding what to discuss in a meeting, planning what to research over the next few weeks, etc. “Rate out of 10 for priority” differs in meaning in each context, but I normally have some easy corresponding gut feel. There’s a surprising amount of value in having any order at all (even an arbitrary one like first-come-first-served!) to break past indecisiveness about where to start. But since I generally have limited time and won’t get to everything, having a rough priority order is a significant upgrade.

I’m experimenting with writing short posts and not overthinking it - let me know how you find this style! I also wanted to get in on Goodhart week before it ended <3

28

3 comments, sorted by Click to highlight new comments since: Today at 4:36 PM
New Comment

See also Final Version Perfected - a very effective task ordering method - but one that does depend on comparing tasks. 

Something I have done in such cases is pretty close to the suggestion: Come up with a rating - but a coarse one, e.g. just low/medium/high - and then only compare the high ones further. The assumption is that the relative priorities of the lower ones might have changed when you get to them. 

Seems potentially useful.  If nothing else, it might be less vulnerable to the failure mode in which having multiple important tasks can make it tempting to do neither one.  

I find that merge sort starts becoming useful for items as few as a deck of cards. But bucket sort is probably better anyways

New to LessWrong?