Comments on: A general model of productivity?

By: On Units of Measure « higherprocess.com Blog

On Units of Measure « higherprocess.com Blog — Wed, 24 Feb 2010 22:49:16 +0000

[...] measuring academic productivity, James Keirstead assembled a “general model of productivity,” following the standard “outputs per unit of input” formula. His units of measure suit [...]

By: Academic Productivity » Who does Google think you are?

Academic Productivity » Who does Google think you are? — Wed, 16 Sep 2009 16:47:57 +0000

[...] One of the themes we’ve been discussing here is the idea that prestige and attention are the main currencies of academia. So it only makes sense that you want your online [...]

By: james

james — Fri, 26 Jun 2009 14:17:35 +0000

Wow – that’s great Benjamin. I’ve started gathering some data and hope to post something in a week or two (I’m going to a conference next week which will mess up the data slightly).

Quick question: have you used any special techniques to assign the 1-10 scores for usefulness and difficulty or are they best guesses?

By: Benjamin Deschamps

Benjamin Deschamps — Fri, 26 Jun 2009 13:40:05 +0000

Here is a bit of data from three days of "work"... I am a Master's student and am just starting to write my thesis, so my work consists of reading, writing, research, procrastination, etc. Day 1 Task | Usefulness | Difficulty | Time | Productivity Planning | 4 | 3 | 15 | 2.4 Trying prod. formulas | 2 | 4 | 120 | 0.7 Solved math question | 1 | 8 | 135 | 0.3 More on prod. formulas | 2 | 4 | 60 | 0.8 Installed soft. imported i | 5 | 4 | 90 | 2.0 TOTAL productivity = 6.3 avg. hourly productivity = 1.3 Day 2 Task | Usefulness | Difficulty | Time | Productivity Planning | 4 | 2 | 15 | 2.7 Fixed comments, form. | 7 | 4 | 90 | 2.7 PDF settings LaTex | 4 | 3 | 45 | 1.9 Wasted time online | 2 | 1 | 45 | 1.2 Wasted more time | 1 | 2 | 30 | 0.6 Continued formulas | 7 | 4 | 150 | 2.5 Unaccounted time | 0 | 1 | 45 | 0 TOTAL productivity = 11.6 avg. hourly productivity = 1.7 Day 3 Task | Usefulness | Difficulty | Time | Productivity Email for meeting | 7 | 4 | 30 | 3.4 Wasted time online | 1 | 1 | 30 | 0.7 Planning | 5 | 3 | 45 | 3.0 Fixed intro, objectives | 9 | 6 | 90 | 3.3 Found and read manual | 5 | 3 | 60 | 2.2 Played with some data | 5 | 4 | 60 | 2.1 Unaccounted time | 0 | 1 | 135 | 0 TOTAL productivity = 14.6 avg. hourly productivity = 2.1 When I look at these results, I agree with the way the formula ranks tasks and total productivity... You can also add a a task with 0 usefulness and difficulty of 1 that accounts for the rest of a workday (wasted time that is not accounted for in these tasks); this then allows you to compute your average hourly productivity for the day. The choice of the log function is based on a few considerations. Without it, the value of productivity is a very small number, difficult to understand and compare. Also, putting work on a logarithmic scale reduces the influence of tasks that take a long time. At the same difficulty and usefulness, a shorter task should be considered more productive. At equal usefulness and equal time, a less difficult task should be considered more productive. These rankings are only achieved if I use the log function. Any thoughts on these results? Have you started collecting data?

By: james

james — Wed, 17 Jun 2009 16:56:19 +0000

Thanks for pointing that out Benjamin. The idea of multiply the attention required by a task and the time it takes (the a*h denominator) is definitely similar to the physics notion of work — which maybe means we’re onto something.

It would be great to hear a bit more about your data collection and experiments too. I’m curious about the choice of the log function.

By: Benjamin Deschamps

Benjamin Deschamps — Tue, 16 Jun 2009 18:14:58 +0000

Quick follow-up, the formula p = u / log(dt) seems to rank task productivity correctly (p is productivity, u is usefulness, d is difficulty and t is time). In my tests, difficulty and usefulness were ranked on a scale of 1-10 and time was in minutes. I am going to try this for a couple of days and see if the results make sense…

By: Benjamin Deschamps

Benjamin Deschamps — Tue, 16 Jun 2009 17:51:18 +0000

Interesting post. What about using a notion similar to the physics notion of “work” as a measure of force per distance? The “work” performed on a task is equal to the difficulty of the task (amount of brain power, concentration, attention, exertion) times the amount of time that that effort is sustained. Then you can either attenuate or amplify this figure according to the usefulness of the task. For example, running 5km and reading a journal article are tasks of the same duration and intensity, but one is (academically) productive and the other is not.

By: james

james — Tue, 16 Jun 2009 08:36:35 +0000

Thanks for the feedback Jose.

I haven’t found any good tracking software yet (though I haven’t really looked either). It would need to be fairly robust though: I’m thinking of a database model where you can enter projects and track related inputs and outputs over time at 15 minute intervals (or less?). Perhaps I’ll try it next week, just with pen and paper and see where I get.

But yes, defining an Atnu seems like a good first step. I think this is a job for AHP too and so we’d need to start by building up a list of tasks that can be pairwise ranked for attention requirements. Again thinking out loud, some examples might include:

- Reviewing a journal article for journal x
- Sketching out the idea for a new journal article
- Preparing a journal quality figure
- Preparing for a one-hour lecture

If people can post these types of tasks as comments, then we can rig up a SurveyMonkey form or something to calculate some Atnu values.

By: Anelly

Anelly — Tue, 16 Jun 2009 08:02:08 +0000

I admire you. I have never read before a so complex post about productivity with formulas and models. Great work James.

By: jose

jose — Mon, 15 Jun 2009 19:32:04 +0000

Killer post. Demonstrates that you can have a formal model (with equations) in a blog and that a post is not necessarily ’soft reading material’. Lots of thought must have gone into this model. As simple as it is, I think it captures a lot of the action. Using attention and prestige (two key concepts we’ve proposed here at AP.com before) seems to be really useful.

I’m up for testing this. Empirically.

Does anyone know of any good software (maybe a website) to track this kind of data?

And, maybe the first step, let’s try to really pin down the value of an Atnu. Reading an article varies in effort (some are more difficult than others), but could be a good proxy since everyone in every discipline needs to do this.

Just thinking outloud…