Oversimplified cause

The missed call

A youth soccer coach tells the parents that the team lost the final because the referee missed a handball in the 85th minute. The team had conceded three goals before then and hadn’t had a shot on target all game.

The clear-cut case. The missed call may have mattered; a goal from that moment might have changed the result. But the loss took everything that happened in the first 84 minutes too. Treating the call as the reason for the loss lets the team skip the parts of the game it could actually work on.

The library’s visitors

A city library’s visits fell by a third over five years. A consultant’s report attributes the drop to the rise of e-books and recommends shifting the print budget to digital. Over the same years, the library cut its evening hours and lost its free parking lot to a construction project.

This comes with numbers and a real trend, which is what makes it persuasive. E-books may well be part of the story. But the report presents one factor as the explanation when at least two others could have kept people away, and its remedy follows from that choice. If evening hours and parking matter, moving money to digital won’t bring those visitors back.

The tomato plant’s “percentages”

A gardening column says a tomato plant’s harvest is “about 60% soil and 40% sunlight,” and advises readers with sunny yards not to worry much about their soil.

This looks more careful than naming one cause, because it credits two. But soil and sunlight don’t add up the way the percentages suggest: a plant in rich soil and no light produces nothing, and so does one in full sun and no nutrients. When each factor is needed for the other to have any effect, there is no fact about what share of the harvest each one “caused”. Mill made the general point: each of the conditions was “equally indispensable to the production of the consequent.” Splitting credit makes sense when causes contribute separately, not when they work only together.

The fallacious version quietly adds a premise nobody established: that the other conditions made no difference, or would have produced B anyway. That’s what licenses conclusions like “remove A and B goes away” or “A is to blame”. Which conditions make up the full set is a separate question, and often a hard one; the error is proceeding as if it had been answered with a single item.

• The sole cause: one contributing factor presented as the only one, as in the missed call. Hurley and Watson’s logic textbook counts this as a variety of false cause under the name oversimplified cause; T. Edward Damer’s textbook calls it causal oversimplification.
• The single-factor theory: explaining a whole class of events by one principle. Mill called this “false theory”, which draws conclusions from “some one agent which is known or supposed to be present, all other co-existing agents being overlooked.” His example was the schools of medicine that traced all disease to a single source, such as “morbid viscidity of the blood.”
• Blame or credit for one person: attributing a group’s failure or success entirely to one member, such as a manager credited with a turnaround that also depended on a recovering market.
• Splitting joint causes into shares: assigning percentages to factors that only work together, as in the tomato example.

Its relatives. The Stanford Encyclopedia of Philosophy reports that Mill also saw post hoc ergo propter hoc as a fallacy that “tends to single out a single cause when there are in reality many contributing causes.” The two often travel together, since the factor singled out is frequently the last thing to happen before the outcome. They are different errors, though: post hoc concludes that something is a cause at all because it came first, while an oversimplified cause may name a genuine cause and go wrong only in treating it as the whole story. Framing a complex outcome as “either A or B” is also close to a false dilemma, when the answer is both, or several things at once.

Naming one cause is ordinary and often exactly right. It’s legitimate when:

• The question asks about one factor. “Would the fire have started without the dropped cigarette?” asks about the difference one condition made, holding the rest fixed. Answering “the cigarette caused it” doesn’t deny that dry weather and oxygen were needed too.
• The factor named is what changed, or what was out of the ordinary. People normally call a match strike the cause of a fire and the oxygen in the air a background condition. Philosophers disagree about whether that distinction reflects anything objective or only our interests, but for practical questions (what went wrong, what to change, who is responsible) singling out the abnormal factor is standard.
• The evidence shows one factor dominates. When removing one factor removes most of the effect, and the factors contribute separately, calling it “the main cause” is accurate.
• The simplification is labeled. “The biggest driver” or “one important reason” claims a share, not the whole.

The test: is this factor being said to have made a difference, or to be the only thing that did?

The fire investigator

After a small grass fire in a park, the fire department’s report states: “Cause: a lit cigarette discarded into dry mulch beside the path.” It doesn’t mention the dry weather, the wind or the oxygen in the air.

The report names one factor out of many that were all needed. But it answers a specific question (what started this fire, out of the conditions that are present every summer day), and it doesn’t claim the other conditions were irrelevant. The legal theorists H. L. A. Hart and Tony Honoré, as Dmitri Gallow’s Stanford Encyclopedia entry reports, suggested that conditions which are normal are treated as background conditions and those which are abnormal as causes. Dry mulch and air are always there; the cigarette was the thing that wasn’t. That’s the what changed condition.

The broken scanner

A delivery company reviews 200 late deliveries from the past quarter and traces 150 of them to a single warehouse, where a broken barcode scanner forced staff to log packages by hand. The review recommends replacing the scanner first, and lists traffic and weather as smaller contributors.

This singles out one cause, but it’s backed by a count showing how much of the problem that cause accounts for, the delays it produced are largely separate from those caused by traffic and weather, and the review names the other factors instead of dropping them. That’s the evidence shows one factor dominates and labeled conditions together.

One cause makes a better story. It’s easier to remember, easier to explain, and it points to a single thing to fix or a single person to blame. Several interacting causes resist all of that.

Mill offered a more specific reason. The factor that gets called the cause is usually the event that completed the set, while the other conditions were states that had been in place for some time. Because nothing else is waited for once that final event happens, “the appearance is presented of a more immediate and close connection between the effect and that one antecedent, than between the effect and the remaining conditions.” In fact, he wrote, that condition “has really no closer relation to the effect than any of the other conditions has.”

Conditions that are always present also go unnoticed, so the factor that stands out gets the credit. And when the outcome is someone’s behavior, people tend to explain it by the person’s character and overlook the situation, a pattern studied as the Fundamental attribution error.

• Ask what else had to be true. Would this factor have produced the outcome by itself, in different circumstances?
• List what changed around the same time. If several things did, the explanation should say why this one matters more.
• Check the remedy against the explanation. If fixing only this factor is expected to undo the outcome, is there evidence the other conditions wouldn’t produce it anyway?
• Be wary of percentages for interacting causes. Ask whether the factors contribute separately, or only in combination.
• Don’t over-correct. Pointing out that there were other causes doesn’t show the named factor didn’t matter. It may still be the one worth changing.