top of page

The -udge Returns: The Kludge!

Excuse me [formal address] do you have time to discuss our Lord and Saviour, the -udge? Today in the -udge saga: the kludge! Now ironically, I think kludge, being kludged and kludging are older than the sludge and the budge, so I feel slightly behind on the times, but nonetheless, here is its dedicated blogpost!


For those of you who are completely not on the same page, behavioural science has had this rather odd tendency, grounded entirely into a love for rhyming, to have phenomena and concepts end with udge. This started of course with the very foundation, the nudge, followed by the sludge (its opposite), the budge (a different variation) and then I’m sure there’s others that just haven’t come to mind yet. However, I was greatly horrified when I realised there was yet another. The Kludge. Mentioned by Robert Metcalfe (yes, him) on the Behavioral Economist Facebook group, as he sent through a screenshot of the paper called Kludged. I was immediately intrigued. Another -udge! Now kludge is an existing word, just like nudge was. A kludge is “an ill-assorted collection of parts assembled to fulfil a particular purpose”, or as a verb: “improvise or put together from an ill-assorted collection of parts.” Now what does this mean for behaviour. Well, in the words of Jeffrey C. Ely (2007): “A kludge is a marginal adaptation that compensates for, but does not eliminate fundamental design inefficiencies. When kludges accumulate the result can be perpetually sub-optimal behavior.” Oh we’re in a for a wild one here!


The paper starts of naming quite a few kludges as seen through both history and nature. Both Microsoft Vista and Emperor Penguins get a good beating, and you start to really get a good feeling for what a kludge looks like. And you also start to realise that they are very common. Maybe a bit too common for comfort… According to Ely there’s three parts to a Kludge:

  1. the system must be increasing incomplexity so that new problems arise that present challenges to the internal workings of the system.

  2. a kludge addresses the problem by patching up any miscoordination between the inherited infrastructure and the new demands.

  3. the kludge itself– because it makes sense only in the presence of the disease it is there to treat– intensifies the internal inefficiency, necessitating either further kludges in the future or else eventually a complete revolution.

Doesn’t this sound familiar?


We move on swiftly from Microsoft and penguins onto our own brains. We are self-centered after all. The brain, ever so unfortunately, is adaptive. Which means its highly kludge-seeking. “An adaptive process is not forward-looking and certainly not governed by dynamic optimization. An adaptive process inherits its raw material from the past, occasionally modifies it by chance (mutation or experimentation),and selects among variants according to success today.”

We’re so screwed…


Now the rest of the paper focusses on building the model and there’s formulas everywhere and the word algorithm gets thrown so often it might as well be the latest Muse album or complete rookies discussing day trading, and somehow, somewhere the Dirichlet family of distributions got involved (Angela Lansbury help me), but the gist of the paper is the following: Suppose μ < 1/6. When q > 0 there is a positive probability that the organism (us) will be forever kludged and thus asymptotically structurally inefficient. Now to give you some context. Ely has adopted “a simple model of mutation and natural selection designed to capture the effects of a general class of adaptive processes.” As mentioned earlier, I will not even attempt to dive into the complexities of the mathematics, you can do that yourself if so inclined. What the paper does is essentially proving that as we are now and have been, we are much more likely to be kludged than to become optimal. In complicated speak: “With probability q, the organism increases in complexity. It keeps the analysis simple to assume that when complexity increases it increases by two, and the two additional computational steps are allocated optimally taking as given the existing allocation. On the other hand, with probability (1−q) the organism does not increase complexity, but some (possibly empty) subset of existing computational steps are re-allocated.” Now we have q explained, let’s attempt to do μ: “There is a fixed mutation probability μ > 0 and each gene (used as an example) is subject to mutation with independent probability of μ.” There we go.


Leaving the mathematics behind, as far as we ever got into them in the first place, it’s interesting to have mathematical proof of the fact that we’ll always be kludged, and that it is very difficult for us to ever become “optimal” given the amount of kludges already in place. It’s not your fault darling, it’s your faulty wiring. This does once more shine a spotlight on the homo economicus, and the debate of rationality as a standard. Personally I don’t find the distinction between rationality and irrationality particularly interesting or useful. To me, it just is. If you want to call a dead fish a helicopter by all means do so, it doesn’t change the awful smell accompanying it. So let’s use this hopelessly complicated article from at least a decade ago to admit to ourselves that we’re kludged. We’ve been kludged for millennia and we’ll continue to kludge on. I find it just takes the pressure off. And for those of you who ever dream of becoming optimal, maybe find a different dream. And for those of you who just want to disentangle the wires and see where they lead, well, welcome to the behavioural sciences!

Behavioural Science

Personal Finance



bottom of page