by Eliezer Yudkowsky Mar 9 2015 updated Apr 2 2015

CICO is a proposed conceptual decomposition of the causes of changes in human body mass, particularly fat content, into the sum of two mostly independent variables for calories consumed by eating and calories expended in exercise.
The implication is that these two quantities can mostly be considered independently (e.g., eating less will not automatically cause your body to expend fewer calories) and that 'a calorie is a calorie' in both CI and CO (there are not some types of calories consumed that are less likely to produce calorie expenditures, nor particular forms of metabolic expenditure that would make people more or less likely to eat more).

Mass-In-Mass-Out analogy

To see why the conceptual decomposition of Calories-In-Calories-Out could be useful, or alternatively useless, independently of arguments about thermodynamics, consider the corresponding conceptual decomposition of Mass-In-Mass-Out, in which changes in body mass are held to be the sum of the amount of mass taken into the body minus the amount of mass that leaves the body.

Mass-In-Mass-Out is even more absolutely true than CICO, since, e.g., calories sometimes leave the human body by excretion or via ketones in urine. However, the corresponding pragmatic advice of MIMO, that someone should inhale less air or exhale more carbon dioxide or drink less water in order to lose fat, is false; inhaling more automatically causes more exhalation, and drinking less water does not avoid the particular kind of mass that contributes to fat storage.

Thus, 'a gram is not a gram' (different types of mass intaken have different effects on long-term mass trends), and it's not possible to intervene on air inhaled independently of air exhaled (the two variables change in lockstep, so it's not useful to think about how to causally intervene on only one of them if you want to change their sum). Thus, although MIMO is thermodynamically true in some sense, it's useless as a causal model or as a qualitative guide to intervening on body mass.

In this pragmatic sense, someone should certainly expect that searching will produce at least some examples of 'calories that are not calories' (the same quantity of CI has different effects on weight) or cases where CI and CO cannot be affected independently. We should also expect that it is possible to find some implications of the 'tautological' part of the formula that are true, e.g., to gain muscle mass it is necessary that you take in extra protein, and protein has calories.

Is CICO a useful model?

The question is to what degree CICO is useful for most people most of the time as a way of considering causal interventions intended to produce weight change - whether a calorie is usually a calorie, whether CI and CO can usually be intervened-on independently, and whether most changes to weight can be interpreted as something that would be expected to affect calorie input xor something that would affect calorie expenditure such that a corresponding intervention on that variable in the opposite direction would prevent the weight change.

Example 1: Imagine an animal whose fat cells behave as a perfect battery, such that all Calories In are immediately stored as fat and all Calories Out are fed by released fat. Suppose that the type of food and the type of caloric expenditure have little effect on how the battery operates, and that current weight has no effect on CI or CO so long as the weight is greater than 0 and less than 100 kilograms. Then so long as total fat is between 0 and 100 kilograms, CI and CO are determined independently and have no effects on each other, and we would expect changes in CI-CO balance to produce corresponding changes in consistent weight trends. CICO would then be a very useful way to think about how to intervene on the animal's weight change, and so long as you understood the effect of an intervention on the calculated quantity for CI, or CO, you would be able to guess its effect on the animal's fat stores.

Example 2: Suppose it is the case for some animal that its CI and CO usually change in lockstep, so that, e.g., its eating more automatically causes it to raise body temperature, move around more, or dump excess calories into urine. Suppose that being fed a particular chemical causes this animal's fat cells to take in more metabolic energy from the bloodstream, while refusing to release stored fat in response to metabolic demand. In response to detected low levels of blood sugar, the animal tries to eat more. Then while this chemical causes the animal's CI to increase, it will not be useful to say that the animal's weight gain was caused by it eating more, and it will not be possible to prevent the resulting weight gain by feeding the animal less. We can further imagine that even trying to simultaneously feed the animal less while stimulating it to more movement fails to produce loss of fat - its fat cells are simply refusing to release triglycerides, and instead continue to take in available metabolic resources from the bloodstream and convert it to fat. Rather than increasing CO, the animal's body temperature drops, or it fails to move when prodded, or it consumes muscle rather than fat as emergency fuel; the animal eventually starves when it runs out of internal protein to consume, with its fat cells still intact and adding triglycerides. CICO remained thermodynamically true, but it was not a useful way to think about how to intervene on this animal's weight changes.

Since CICO is conceptually a piece of pragmatic advice for thinking about the causes and intervention points on human body mass, especially fat stores, the question is whether the average human animal behaves more like Example 1 or Example 2, or rather, where the average human animal falls on the spectrum between these two extremes. CICO could remain a pragmatically useful way to think about fat gain even if, e.g., some stimuli caused fat cells to try to store more fact, if it remained true that trying to intervene on decreasing caloric intake or increasing exercise or both was able to prevent the resulting weight gain.

Evidence and Arguments