2.1Knowing the end changes the reading
Fischhoff's paper of 1975 is the starting point for the first position. Its reading note says that people told how events turned out judge that outcome more likely than they would have without knowing it, and that they do not realise the knowledge changed their judgment (Fischhoff 1975). The second half of that claim carries the weight. A reader who knows the end is not only more confident; the reader cannot tell that the confidence came from the end. Fischhoff called the effect creeping determinism.
Baron and Hershey's paper of 1988 makes a related but distinct claim. The reading note says that people rate the same decision, made on the same information, as better when its outcome was good than when it was bad, so that the outcome colours the judgment of the decision (Baron and Hershey 1988). Fischhoff's claim is about the probability assigned to the outcome. Baron and Hershey's is about the quality assigned to the decision. The two can come apart. A partner might accept that a storm was unlikely and still say that sending the ship was a bad decision, because it was sent and lost.
Taleb's The Black Swan (2007) holds that rare events of extreme impact are explained, after the fact, as if they had been predictable (Taleb 2007). The claim moves from what a person believes about probability to what is said about the past, and that move is the one Mackay's chapter illustrates.
2.2Outcomes, counted, are the test
The second position is not a single school, and its two works do not, in their notes, argue against Fischhoff or Baron and Hershey; the opposition follows the parent study's plan for its chapter on judging decisions. Brier's paper of 1950 proposed a way to score probability forecasts against what happened: the mean squared difference between forecast and outcome, a score that a forecaster cannot improve by stating anything but the probability believed (Brier 1950). The score judges forecasts against outcomes. Across many forecasts, a single bad outcome after a forecast of a small probability weighs little, while a long series of forecasts that do not match what happened weighs a great deal.
Tetlock's Expert Political Judgment (2005) is a long study of experts' forecasts on political and economic questions. The reading note says the forecasts were poorly calibrated and often no better than simple rules, and that experts who drew on many ideas and revised readily, the foxes, did better than those committed to one big idea, the hedgehogs (Tetlock 2005). The study shows what judging by outcomes can do when the outcomes are many. Its material is forecasts, and forecasts are made before the events they concern.
So the second position does not deny hindsight. It moves the judgment from the single case to the series, and it depends on something written before the outcome, the forecast, as much as the first position does.
2.3The record before is not innocent
The third position comes from organisational sociology, and it questions the document on which the other two rely. Power's The Risk Management of Everything (2004) argues, as its note gives it, that risk management spreads into all organisational activity and turns towards secondary risks, the reputational and legal risks to the organisation and its managers, so that defensive documentation can take the place of judgment (Power 2004). Applied to a record of reasons, the point cuts close: a reason can be set down for a later inquiry rather than for the decision, so that, if things go wrong, the author can show it was set down.
Power's later Organized Uncertainty (2007) describes risk management as a way organisations turn uncertainty into objects they can manage and account for (Power 2007). A log of reasons is such an object: it makes a judgment reviewable, and it gives the judgment the shape the log's form allows.
Clarke's Mission Improbable (1999) is about a different kind of document, plans for catastrophes that cannot be controlled. Clarke calls them fantasy documents: they could not work as written but serve to persuade others, and their authors, that the risk is under control (Clarke 1999). The idea is borrowed here for a narrower purpose. A record of reasons can be written to persuade as well as to inform. A merchant who knows the partner will read the log has reason to write a log that will read well.
2.4Episodes told as sequences
One more work bears on how Mackay's chapter is read. Kindleberger's Manias, Panics, and Crashes (1978) is, as its note gives it, a history of financial crises in a recurring sequence of displacement, boom, euphoria, distress and panic, and of the lender of last resort (Kindleberger 1978). A sequence of that kind is a way of telling an episode once its end is known: each stage is named by what came after it. That is no charge against the sequence; any account of a crisis is written after it. It marks the sequence as a narrative form, of which Mackay's chapter is an earlier and cruder instance.
2.5The positions side by side
The table summarises the three positions by the questions each answers. The words in the cells are the article's, not the authors'.
| Question | Knowing the end distorts (Fischhoff; Baron and Hershey; Taleb) | Outcomes counted are the test (Brier; Tetlock) | The prior record is not innocent (Power 2004; Clarke) |
|---|---|---|---|
| What is the danger? | The outcome makes itself look likely and the decision look bad | Judging from one case, or from stories rather than scores | Records written to defend or persuade |
| What is the remedy? | Judge from what was known at the time | Score many forecasts against what happened | Read the record as an act with its own purpose |
| What does it need written before? | The information available, kept separate from what came later | The forecast, stated as a probability | Nothing more; it asks who wrote the record and for whom |
| What can it not do? | Say whether the decision was in fact good | Judge a single decision | Supply a standard of judgment of its own |
The table hides how much the first two positions share. Both need something written before the outcome. Fischhoff and Baron and Hershey need it so that the judge can recover what was known; Brier and Tetlock need it so that there is a forecast to score. The third position doubts that what is written before is a neutral trace of the decision. That doubt applies to both of the others.
2.6What a narrative source can add
None of these works, in its note, discusses how a collapse is told. Mackay's chapter is a telling made long after the event, by an author who knew the outcome, for readers who knew it too. If hindsight and outcome bias are real features of judgment, they should be visible in it. Mackay also quotes documents written at different times, a letter from the month of the fall, a committee's reports, a historian's memoir, which show how a telling uses material written before and after the turn.