What’s the origin and precise meaning of this phrase? I gather it basically means “hard to learn”. But why the (presumably) graphing-related imagery, and did the phrase arise out of an actual casestudy of intelligence or learning?
Also, while I’ve heard it used most often to describe a task, when applied to a person (“he has a steep learning curve”) I’ve heard it used alternately to mean that the subject is sharp-as-a-tack or slow-on-the-uptake. Which one is correct?
Doubt there’s a traceable scientific origin. I always thought it was just another bullshit buzzword.
Always heard it applied to subject matter, not people.
I’d always understood it to mean the requirement to go from virtually no knowledge of something to mastery very quickly. For example, saying that a job has a steep learning curve means you don’t get a lot of time to learn the ropes.
I’ve always wondered about this too.
If ever a “learning curve” were plotted, you would expect it to be “amount learned” against “time”, since “time” is usually plotted on the horizontal access. For a given subject, a “steep” curve thus plotted would indicate that you acquire a lot of knowledge about the subject in a short time. Thus, the imagery could be that you are forced to cram a lot of information in a short period.
However, it has now been hijacked as a corporate buzz-phrase. I think what happened was that someone took the “steep” metaphor too literally, and, applying the real-world concept of “steep”, took it to mean that it’s difficult to learn a given subject - like climbing a “steep” hill. Of course, graphical curves aren’t subject to gravity so this is meaningless, but there’s a lot of ignorance out there.
Here’s a reference to an application of the learning curve in the aircraft industry in 1936, implying that the first use might have been earlier.
Here’s Wikipedia’s article on learning and experience curves, indicating the learning curve effect was first quantified in 1925.
Wikipedia actually has a pretty good introductory article on the concept of the learning curve. Because I can’t put it any more succinctly, I’ll quote from the article:
The learning curve effect and the closely related experience curve effect express the relationship between experience and efficiency. As individuals and/or organizations get more experienced at a task, they usually become more efficient at them. Both concepts originate in the old adage, “practice makes perfect”.
So, an experience vs. efficiency curve generally looks somewhat logrithmic; a little bit of experience doesn’t produce a proportional gain in efficiency over base neophyte (the tangent is very “steep”–over unity–initially), but once you get to a certain level of experience the curve “breaks over” (where the tangent becomes fractional) and you become more adept at figuring out how to anticipate and solve problems and where to find data.
In manufacturing, the notion of a learning curve has been used for a long time–Robert McNamara, later Secretary of Defense, introduced it to Ford Motor Company when he became part of their nascient quality department–to quantify the process of “getting the kinks out”. It is often incorrectly conflated with efficiencies of scale (which come about due to cost/labor reductions of bulk purchasing and automation).
Here’s[sup]*[/sup] a funny example of “learning curve”: When physicist Richard Feynman started doing research on the properties and behavior of superfluid helium (He II) in the early Fifties, he spent a lot of time looking in journals, talking to experimentalists, and otherwise becoming acquainted with the state of knowledge regarding that substance. He came up with a novel theory of its behavior, but despite extensive efforts couldn’t figure out the details of the thermodynamic functions at the phase transition. He presented a paper on his theory at a conference in Japan in 1951, apologizing for his inability to address this particular detail. Another, more experienced researcher in the field. Lars Onsager, spoke up and said, “Mr. Feynman is new in our field, and there is evidently something he doesn’t know about it, and we ought to educate him.” Feynman was mortified and was prepared for a scathing attack on his ignornance, but Onsager continued, “so I think we should tell him that the exact behavior of the thermodynamic functions near a transition is not yet adequately understood for any real transition in any substance whatsoever.”
In other words, if Feynman had known what other physicists involved in thermodynamic research of material phase transitions all knew he wouldn’t have spent much time or felt the need to apologize. Heh.
Stranger
I’m currently reading Perfectly Reasonable Deviations From the Beaten Track, a collection of correspondence from and to Feynman, and he brings this up in a letter to Onsager’s biographer. He appears to have later retooled this passage for one of his own autobiographical collections of anecdotees (“Surely You’re Joking, Mr. Feynman?”* I believe).
The “steepness” comes from the tangency (the instantaneous ratio of rise over run at some point on the curve.) If you can quantify experience (say, units produced) with some measure of quality or efficacy (number of units rejected, amount of time to produce x units, et cetera) then you can fit a curve, and generally this curve follow a logrithmic distribution (at least, up to a point, and in absence of other extenuating factors).
So, the term isn’t meaningless or metaphorical, but no doubt “learning curve”, like “Six Sigma”, has been hijacked by MBAs who aren’t capable of balancing a checkbook much less understanding statistics or calculus, and who think the “TAN” button on their HP business calculator will give them skin cancer.
Stranger
I does seem to me that DarrenS is correct in that the meaning in popular usage has been essentially reversed (or perhaps has gone off on a tangent ;)). “Steep learning curve” in general usage seems to typically mean that something is hard to learn, or that somebody has a hard time learning something - rather a bad thing. In the original meaning, a steep learning curve would mean that a little learning results in rapid benefits, a good thing.
I suppose. In dealing with pilot production of new product lines (I’m a mechanical engineer working in design and structural analysis and have had to work with manufacturing departments in the past) we’ve always plotting something like number of defects or labor cost per unit or per operation, so a “steep” curve indicated many initial problems; with an increasing number of units, the design and manufacturing process problems were (ideally) ironed out, resulting in a “flatter” curve with respect to successive production.
It just depends on how you orient and label the axes, I suppose.
Stranger