Comments on: One reason why plans are good

By: Brad Hutchings

Brad Hutchings — Sat, 16 Jun 2007 21:46:22 +0000

I agree with Arnold Kling above. I recently found myself in a phone meeting with my collaborators and a guy helping us with a marketing project. Some background… If you were to watch me and my collaborators work on the project we’ve been doing for four years now, you would think that from day to day or week to week, we are entirely random, but over the long haul, a clear direction is present. Such is how R&D intensive ventures tend to work, in my experience. We need ongoing micro failures as well as ongoing micro successes to learn what we can pull off and what we can’t. So the marketing guy has a goal in mind for a couple months out and suggests we come up with a spec and a project roadmap as a first step. When I hit him with “we don’t do specs”, his question (asked in a tone of total disbelief) was “well then how do you figure out what you need to do?”. At any rate, two months later, we had just what he needed for the particular task and it didn’t resemble anything we talked about at first. Go figure.

Where I agree with Kling and would extend his thought is that planning an ongoing part of the process, not a prelude to the process. Software people figured this out 20 years ago, when they started looking at the “spiral model” of rapid prototyping and revision to finished product as an alternative to the traditional “waterfall model” of spec –> design –> implementation –> test –> maintenance. It takes a different mindset.

By: Hopefully Anonymous

Hopefully Anonymous — Thu, 14 Jun 2007 22:01:34 +0000

Stuart, so … what? I’m still interested in the science of dividing up planning for a goal and assessing how much planning one should engage in to accomplish that goal (call it planning if you like, too, or call it smurfy -whatever). I’m sure there’s a recursive element -should one assess how much assessing one should do? I’m curious who has teased out the nuances of this in detail. I imagine people studying business management and artificial intelligence related fields would need to do so.

By: Stuart Armstrong

Stuart Armstrong — Thu, 14 Jun 2007 19:11:53 +0000

Prior to engaging in planning, one should assess (by bayesian analysis, etc.?) whether or not the planning process itself is likely to reduce one’s resources more than one is likely to gain by the goal accomplishment.

That sounds suspiciously like… planning.

By: Michael M. Butler

Michael M. Butler — Wed, 13 Jun 2007 16:50:59 +0000

Sorry, I realize I drifted the topic.

More directly on point for Andrew: It seems that to plan is to decide. “Decide” has as its root the Latin for “to cut away”. You’re pruning choices. Even if all you’re deciding is the first step of a plan, it’s going to reshape the steps that follow, or their likelihood/degree of difficulty.

And to consciously undertake to do _something_, it seems that deciding precedes acting, even if the decision is not fully conscious enough to be considered the result of a plan.

All this is modulo determinism-behaviorism, of course, and leaves out whatever “habit” is. There’s a strange loop there of cultivating habits that I’ll mention and then ignore.

In a literal vein: carving granite is different from carving clay. A clay “target model” for a granite sculpture is a plan that itself was both planned and re-evaluated as it developed. And the granite might still turn out to have its own surprises (hidden discontinuities or colors, etc.) that will mean the exact plan will need alteration to fit the revealed truth.

I apologize if I’m stating the obvious.

By: Michael M. Butler

Michael M. Butler — Wed, 13 Jun 2007 16:35:52 +0000

I’ve just recently been reminded by circumstance (moving to a new living space) that I’m a good improviser. In many ways, microplans and OODA loops that include planning or pre-planning (practice, “being prepared”) are probably involved. But…

I have a historic tendency to believe I have to “have a Plan” in many circumstances, and that can lead to deadlock when reality diverges. For me, under the right kinds of stress, the desired flexibility in approach is easier come by.

I know (I believe) there are people who lock up under stress. And of course that’s happened to me a couple of times. But more often than not “hero mode” acts as a sort of lubricant for me. Not bragging, just using shorthand.

So: is this all attributable to microplanning and preplanning? Or does it have to do with the quality of attention and intellect I bring to the task when time is of the essence and my perfectionistic “plans must be perfect or I’m an idiot” neurosis is subdued by circumstance?

Put another way: impulse control is most necessary when your impulses are badly biased. Discuss?

Mike

By: LP

LP — Wed, 13 Jun 2007 03:12:12 +0000

Eliezer draws attention to an important feature of plans: they iterate. A good planner has to periodically change direction, adapt the plan to newly acquired knowledge about the goal. A plan that doesn’t do this — one that’s ‘written in stone’ — is a bad plan, unless the planner knows everything there is to know about the universe (c.f. God). The ideal plan would be one that continually adjusts itself to new information, making constant (very small) changes in direction.

By: Tony K

Tony K — Wed, 13 Jun 2007 02:04:07 +0000

I see a lot of confusion between planning, goal-setting, and forecasting.

Forecasting accurately is impossible except maybe in the very short or very long term. You still at some point need to come up with a number. Forecasting is best done using probabilistic techniques and reporting events as probabilities.

Goal-setting is important and helps people keep focused, but goals are not good forecasts.

Plans set up a sequence of events or actions that can be taken towards achieving those goals. Plans, like forecasts are never accurate, but they do help clarify your path.

http://kenckar.blogspot.com

By: ed

ed — Tue, 12 Jun 2007 20:54:29 +0000

There is some common sense that planning is good and necessary. I wonder about different aspects in life some of which are easy to plan while others almost require spontaneous action. This also leads to how we are defining things. If I need to accomplish something in the next 30 seconds and those seconds are spent using 1.0 seconds worth of planning and implementation doing 30 consecutive things, most of us would look at that as being spontaneous rather than planned and yet really each action was planned in your brain instantaeously. Compare that to viewing an implementation that took 30 days instead of 30 seconds.

Also, hindsight is often a problem. When goals aren’t met, we often blame lack of planning. When goals are met even without a “plan” we congratulate someone who is “good on their feet”

I’ll use basketball as an analogy. A star player on a team has been taught plenty of plans for the teams objective to score a basket. He has years of experieince to draw upon. He takes the basket ball and through an enormous amount of skill drives the ball down court and dunks it over the opposition. Was there a plan here or was it spontaneous. Does it matter if he studied the film the night before and recognized the opposing team is slow on defense on the left side after scoring a 3-pointer? As opposed to his nearly instant decision to see a weakness in the defense the second he touched the ball.

Lets also relate this to overplanning. Besides a war plan, the classic case is the business plan. Option A and Option B may actually meet the goal but what if that planning (and for that matter the goal setting) causes the business to miss the boat entirely. I am sure nearly every extinct or bankrupt business had a well defined plan and goal. This doesn’t mean the plans were worth less nor can we suggest that not planning would have succeeded. Instead we can categorize types of planning for different types of scenarios. From rigid step by step planning to necessary and encouraged free wheeling. When is one better than the other. We can go with evolution on that one and simply say – when they succeed.

By: Kyle

Kyle — Tue, 12 Jun 2007 19:35:36 +0000

“In preparing for battle I have always found that plans are useless, but planning is indispensable”

–Eisenhower

One of my favorite quotes, and in the same direction as Mr. Kling’s comment

By: Hopefully Anonymous

Hopefully Anonymous — Tue, 12 Jun 2007 17:16:18 +0000

Eliezer,
You make a good point, except I think it should be located prior to active planning. Prior to engaging in planning, one should assess (by bayesian analysis, etc.?) whether or not the planning process itself is likely to reduce one’s resources more than one is likely to gain by the goal accomplishment. If it is, it’s more rational not to plan. And there seem to be degrees to all this: how much planning one should do, etc. I imagine this has been studied in depth, because I think it would be important both for business management, and for creating AI systems, etc.

Also, I think an alternative to planning, other than random behavior, is default behavior.