Recommendations vs. Guidelines
post by Scott Alexander (Yvain)
Medicine loves guidelines. But everywhere else, guidelines are still underappreciated.
Consider a recommendation, like “Try Lexapro!” Even if Lexapro is a good medication, it might not be a good medication for your situation. And even if it’s a good medication for your situation, it might fail for unpredictable reasons involving genetics and individual variability.
So medicine uses guidelines – algorithms that eventually result in a recommendation. A typical guideline for treating depression might look like this (this is a very over-simplified version for an example only, NOT MEDICAL ADVICE):
1. Ask the patient if they have symptoms of bipolar disorder. If so, ignore everything else on here and move to the bipolar guideline.
2. If the depression seems more anxious, try Lexapro. Or if the depression seems more anergic, try Wellbutrin.
3. Wait one month. If it works perfectly, declare victory. If it works a little but not enough, increase the dose. If it doesn’t work at all, stop it and move on to the next step.
4. Try Zoloft, Remeron, or Effexor. Repeat Step 3.
5. Cycle through steps 3 and 4 until you either find something that works, or you and your patient agree that you don’t have enough time and patience to continue cycling through this tier of options and you want to try another tier with more risks in exchange for more potential benefits.
6. If the depression seems more melancholic, try Anafranil. Or if the depression seems more atypical, try Nardil. Or if your patient is on an earlier-tier medication that almost but not quite works, try augmenting with Abilify. Repeat Step 3.
7. Try electroconvulsive therapy.
The end result might be the recommendation “try Lexapro!”, but you know where to go if that doesn’t work. A psychiatrist armed with this guideline can do much better work than one who just happens to know that Lexapro is the best antidepressant, even if Lexapro really is the best antidepressant. Whenever I’m hopelessly confused about what to do with a difficult patient, I find it really reassuring that I can go back to a guideline like this, put together by top psychiatrists working off the best evidence available.
This makes it even more infuriating that there’s nothing like this for other areas I care about.
Take dieting. Everybody has recommendations for what the best diet is. But no matter what diet you’re recommending, there are going to be thousands of people who tried it and failed. How come I’ve never seen a diet guideline? Why hasn’t someone written something like:
1. Try cutting carbs by X amount. If you lose Y pounds per week, the diet is working. If not, you’re probably resistant to cutting carbs because [two hours of mumbling about insulin] and you should move on to the next tier.
2. Try cutting fat by X amount. If you lose Y pounds per week, the diet is working. If not, you’re probably resistant to cutting fat because [two hours of mumbling about leptin], and you should move on to the next tier.
And so on until Step 7 is “get a gastric bypass”.
I agree nobody can ever make a perfect algorithm that works for all eventualities. But still. Surely we can do better than “Try the Paleo diet! I hear it’s great!”
What information do guidelines carry beyond a recommendation?
First, they have more than one recommendation. It may be that the Paleo diet is the best, but the guidelines will also include which is the second-best, third-best, et cetera.
Second, because they have more than one recommendation, they can tailor their recommendation to your specific circumstances. The person with depression and comorbid anxiety may want to start with Lexapro; the person whose main symptom is tiredness may want to start with Wellbutrin. Since I love bread, does that mean I should avoid carb-cutting diets? Does that mean it’s extra-important that I cut carbs? Does it not matter, and really it depends on whether I have a family history of diabetes or not?
Third, they acknowledge that some people might need more than one recommendation. If you hear “try the Paleo diet”, and then you try it, and it doesn’t work, you might believe you’re just a bad dieter, or that all diets are scams, or something like that. Guidelines implicitly admit that everyone is different in confusing ways, that something that’s expected to work for many people might not work for you, and that you should expect to have to try many things before you find the right one.
Fourth, because they admit you may need to try more than one thing, they contain (or at least nod at) explicit criteria for success or failure. How long should you try the Paleo diet before you decide it doesn’t work? How much weight do you need to lose before it qualifies as “working”? If it’s been three months and I’ve lost four pounds, should you stick with it or not?
Fifth, they potentially contain information about which things are correlated or anticorrelated. The depression guidelines make it clear that if you’ve already tried Lexapro and Zoloft and they’ve both failed, you should stop trying SSRIs and move on to something with a different mechanism of action. If I’ve tried five carb-cutting diets, should I try a fat-cutting diet next? If I hate both Mexican food and Chinese food, is there some other category of food which is suitably distant from both of those that I might like it? Guidelines have to worry about these kinds of questions.
My impression is that once you understand a field really well, you have something like a Guideline in your mind. I think if nobody had ever written a guideline for treating depression, I could invent a decent one myself out of everything I’ve pieced together from word-of-mouth and common-sense and personal experience. In fact, I think I do have some personal guidelines, similar to but not exactly the same as the official ones, that I’m working off of without ever really being explicit about it. Part of the confusion of questions like “What diet should I do?” is sorting through the field of nutrition until you can sort of imagine what a guideline would look like.
So why don’t people who have more knowledge of nutrition make these kinds of guidelines? Maybe some do. I can’t be sure I haven’t read dieting guidelines, and if I did I probably ignored them because lots of people say lots of stuff.
But I think that’s a big part of it – making guidelines seems like a really strong claim to knowledge and authority, in a way that a recommendation isn’t. Some idiot is going to follow the guidelines exactly, screw up, and sue you. I just realized that my simplified-made-up depression guidelines above didn’t have “if the patient experiences terrible side effects on the antidepressant, stop it”. Maybe someone will follow those guidelines exactly (contra my plea not to), have something horrible happen to them, and sue me. Unless you’re the American Psychiatric Association Task Force or someone else suitably impressive, your “guidelines” are always going to be pretty vague stuff that you came up with from having an intuitive feel for a certain area. I don’t know if people really want to take that risk.
Still, there are a lot of fields where I find it really annoying how few guidelines there are.
What about nootropics? I keep seeing people come into the nootropics community and ask “Hey, I feel bad, what nootropic should I use?” And sure, eventually after doing lots of research and trying to separate the fact from the lies, they might come up with enough of a vague map of the area to have some ideas. But this is an area where “Well, the first three things you should try for anxiety are…” could be really helpful. And I don’t know of anything like that – let alone something that tells you how long to try before giving up, what to look for, etc.
Or let’s get even broader – what about self-help in general? I don’t really believe in it much, but I would love to be proven wrong. If there were a book called “You Are Willing To Devote 100 Hours Of Your Life To Seeing If Self-Help Really Works, Here’s The Best Way For You To Do It”, which contained a smart person’s guidelines on what self-help things to try and how to go about them, I would absolutely buy it.
Comments sorted by top scores.
comment by MalcolmOcean (malcolmocean) ·
2018-04-13T17:11:35.857Z · LW(p) · GW(p)
It's not a full-on “You Are Willing To Devote 100 Hours Of Your Life To Seeing If Self-Help Really Works, Here’s The Best Way For You To Do It”, but I did start working on a flowchart to respond to people who asked me for productivity advice. Many people wanted a list of top tips or something, but this is like giving a list of the top anti-depressants: nearly useless. You don't need to read a list of 10 ideas, you need to do a list of 1 idea.
The basic structure of what I got asks "Have you tried the pomodoro technique?" and makes a recommendation for giving it a serious try if you haven't gotten it to work yet. Then if the person has already had major success with the pomodoro technique, or if they've deemed that it won't work for them, it has a few other next major questions. Very much a Version 0.2; I haven't gotten around to expanding it but it feels like it would be worth doing.
Here's an email newsletter that's based on this flowchart (due to lack of limitations in interactivity in email, it doesn't have the full detail of troubleshooting why the pomodoro technique might not work for someone the first time)
comment by romeostevensit ·
2018-04-13T16:31:54.992Z · LW(p) · GW(p)
Because people don't arrive at an explicit model of the thinking process that generates process models until late in their career typically. Whitehead in his later years founded the Center for Process Studies. Gendlin eventually wrote A Process Model. Douglas Englebart founded the Bootstrap institute in his later years. There are other examples I'm not remembering off the top of my head.
A contrast for illustration: Boyd wrote about the OODA loop in his late 40's but never seemed to make the next meta level jump up to trying to instill the kind of reasoning that generated the OODA loop (or EM theory for that matter) pedagogically.
Process models seem very similar to the field of phenomenology in that every few years someone independently reinvents it and creates their own vocab and models. Then they turn it into a management course. So it doesn't seem to advance much. (IIRC Korzybski of 'The Map is not the Territory' fame, basically LW in the 1920's went this route as well). It's a shame because when people do try to do something on this level with a field the results are often interesting. Gendlin came up with Focusing by doing process modelling of psychotherapy before making the jump later to explicitly generalize the mental motion.
Also, it often seems that the people capable of thinking useful thoughts in this area are all in math and ignore practical applications.Replies from: MattG2
↑ comment by MattG2 ·
2018-04-13T23:07:18.922Z · LW(p) · GW(p)
Boyd wrote about the OODA loop in his late 40's but never seemed to make the next meta level jump up to trying to instill the kind of reasoning that generated the OODA loop (or EM theory for that matter) pedagogically.
This is exactly what he did with "The Discourse on Winning and Losing"
Boyd is one of my favorite examples of a great process modeler and meta-level thinker because he did it at every level of his career:
Process modelling of why he was such a good fighter-pilot led to EM theory.
Process modelling of why EM theory worked led to OODA loop.
Process Modelling of how he kept doing great process modelling (including the OODA loop) led to "Destruction and Creation" and Process modelling of how the OODA loop worked led to "A Discourse on Winning and Losing"Replies from: romeostevensit
↑ comment by romeostevensit ·
2018-04-14T01:06:38.716Z · LW(p) · GW(p)
oh awesome I wasn't actually familiar with that. So it fits the pattern of happening later in most careers (dated 1987? he would have been 60)Replies from: MattG2
↑ comment by MattG2 ·
2018-04-14T01:44:02.413Z · LW(p) · GW(p)
I think this is actually a general pattern that happens in most knowledge worker careers, not only late in careers. Certainly when I was a career coach one of the key things I did to help people move up in their careers was to help them move a level up in their thinking.
I think one of the reasons that the particular meta-level up move that you're talking about happens late in careers is that at that point people who are at the top of their careers basically don't have another meta-level up they can move to understand their field - they've already made that move. So the only meta-level they can move to next is to apply the move to itself.
comment by timtheenchanter ·
2018-04-13T19:07:25.842Z · LW(p) · GW(p)
I've been thinking about writing a step by step guide like this for weight maintenance for a few weeks.
Weight maintenance requires long-term adherence. This plan is tailored for practicality and ease of compliance. Notice that each step has concrete goals that you either comply with or not, which makes it easy to assess whether you are on track. The steps are ordered for [efficacy/difficulty].
- body scale
- food scale (if you get to step 4)
- 8 food containers that can hold ~ 250mL or ~ 1 cup (if you get to step 4)
For the scales, have easy to replace batteries (i.e. AA, AAA, not lithium coins that you never can find at the store).
Start at step 1. Maintain a step for 2 weeks and assess if you're losing weight or maintaining an acceptable weight. If yes, then stay on that step. If not, keep doing what you are doing, and also add the next step. Once you reach your target weight, you can assess for two weeks and pull back one step at a time every 2 weeks. Follow this as written for best results, however, for ease of compliance:
- You may skip any 1 step except #1
- You may also switch a step up or down by 1 spot except #1
- Step #6 is required for subsequent steps.
1) Weigh yourself every morning.
2) Stop drinking anything containing sugar (including juice) or alcohol.
3) Include 4 or more eggs with breakfast (280 cal), however you want to prepare them. If you really dislike eggs, 2 scoops of whey protein (~240 cal) can be substituted.
4) Any time you are going to eat (any meal, any snacks, a single cookie, anything), you have to eat one of the following first if it's been more than 1 hour since you have:
- 100 g fresh apple (52 cal)
- 100 g fresh strawberries (33 cal)
- 100 g fresh cherries (63 cal)
- 100 g fresh peach (40 cal)
- 100 g fresh pear (57 cal)
- 100 g fresh blackberries (43 cal)
- 100 g fresh raspberry (53 cal)
- 100 g fresh blueberry (57 cal)
- 100 g fresh baby carrots (41 cal)
- 100 g fresh cucumber (16 cal)
- 100 g fresh celery (16 cal)
- 100 g fresh or steamed spinach (23 cal)
- 100 g fresh or steamed broccoli (34 cal)
These are selected for palatability, cost, ease of use (easy to prepare, no cooking = less excuses). Prepare 100 g containers of the above each night for the next day. Balance fruit and vegetable in equal amounts.
5) Weight training:
- Learn to deadlift with proper form (Youtube videos are probably enough), deadlift 5 sets of 5 reps 1 day per week. Add 5 lbs any time you successfully finish 5x5. A hex bar plus plates don't take up too much room so this can be done at home if you have a ground floor and space available.
- Do push-ups of the best quality you can for 5 sets of 10 once a week. Improve the quality as you get stronger.
6) Count calories using the food scale and published nutritional information on packages / restaurants. After doing this a while, you may be able estimate (be careful though).
7) Limit daily calories to 15.0x your bodyweight in lbs.
8) Limit daily calories to 12.5x your bodyweight in lbs.
9) Limit daily calories to 10x your bodyweight in lbs.
10) Still not working? Check your math. Are you really counting everything? Are you weighing it? Are your units correct (grams / kg / lb / serving sizes)?Replies from: romeostevensit
comment by MattG2 ·
2018-04-13T23:34:22.017Z · LW(p) · GW(p)
There are people who have done this in a variety of fields but they seem to be largely niche. One reason that this may be is that most medicine literally has magic pills that allow you to apply each of the solutions, whereas most other effective process models require a deep understanding of each step to apply them.
For instance TRIZ is an attempt to create a clear process model for "How to Come up With Creative Solutions as an Engineer. " THIS is one process model for TRIZ, and each step in that process model would expand to a diagram that's at least as complicated.
This is an example of one teeny tiny part of the "How to Actually Do Self Help" process model in my head, and each part itself requires deep understanding and background to make any sense.
comment by romeostevensit ·
2018-04-13T16:39:40.855Z · LW(p) · GW(p)
An attempt to instill this that might click for some in the same way that lambda calc first clicks for some, applying TAPs to TAPs causes it to click for some and applying the OODA loop to the OODA loop made it click for me:
Set a five minute timer and apply genre savviness to genre savviness.
comment by MDMA ·
2018-04-13T16:26:09.122Z · LW(p) · GW(p)
I am aware of the "Things that sometimes help for anxiety/depression" list of recommendations that you have made in the past. However, you mention having implicit guidelines for depression, and presumably, anxiety. A blogpost making those guidelines explicit would be greatly appreciated!
comment by Hazard ·
2018-04-13T14:54:40.929Z · LW(p) · GW(p)
Sometime recently I noticed that most of the systems I use to organize my life had to go through multiple serious revisions before they actually became useful. I agree that what would be super useful to read would be something about the process of finding/creating particular systems that work best for you. I haven't really seen any productivity/self-help stuff that wasn't talking about their own particular system. At some point in the nebulous future I intend to write such a guide.
Also, resolving disagreements is something I'd like to see a guide for. LW has a lot of great recommendations for how to do so, but I haven't read a "Here's everything together" guideline.Replies from: romeostevensit, Abysmal Truth
↑ comment by romeostevensit ·
2018-04-13T16:36:46.762Z · LW(p) · GW(p)
Feynman: "He opened a fresh notebook. On the title page he wrote: NOTEBOOK OF THINGS I DON'T KNOW ABOUT. For the first but not last time he reorganized his knowledge. He worked for weeks at disassembling each branch of physics, oiling the parts, and putting them back together, looking all the while for the raw edges and inconsistencies. He tried to find the essential kernels of each subject."
Workflowy is stupendous for this due to how easy it is to rearrange your ontology on the fly. I tried something similar to the above in several areas and it caused a level up. Not just in the object level areas I tried it in, there was a level up in the meta process.Replies from: Hazard
↑ comment by Hazard ·
2018-04-13T18:50:46.164Z · LW(p) · GW(p)
I was thinking of "systems I use to organize my life" very broadly, not just in terms of building one's ontologies (though that's also something I've been thinking about lately, and I'm going to look into workflowy). Things like how have I designed the basic structure of my days, how does my calendar get used, in what contexts will I take notes, how do I manage habits.
A blanket productivity recommendation might be, "Implement GTD". I'm imagining a guide that would offer a process that one could use to determine what parts of their life could benefit from systematization, and how to design, test, and tweak systems so that you end up with one's that work for you.Replies from: romeostevensit
comment by Yannick_Muehlhaeuser ·
2018-04-14T11:31:44.273Z · LW(p) · GW(p)
A Hansonian explanation fo this may be that, say when it comes to dieting, people claim to want science-based, honest guideline to help them loose weight but actually just want to find some simple diet or trick that they need to follow.
Constructing and elaborate guideline might be something that a publicly funded organisation may be able to do, but someone who wants to make money probably won't do that, because it would likely not sell to well.