====== Accident fallacy ====== A [[index|fallacy of generalisation]], that involves applying a general rule to a specific case where an exception is warranted. For example: > It is illegal to injure a human being. > Therefore, it is also illegal for a surgeon to perform emergency surgery. While such a general rule is justified in principle, specific circumstances may necessitate deviating from it. An "emergency operation" – that is, a surgical procedure intended to save a person’s life – is certainly an example of such an exception. ===== Names ===== ==== Alternative names ==== * Fallacy of [the] Accident * Riding on principles * Secundum quid (et simpliciter) * A dicto simpliciter (ad dictum secundum quid) ==== Notes on the name ==== The term "accident", as used in the name of this fallacy, comes from the Latin word "accĭdens ", which refers to a //non-essential// aspect of something. It can be understood as the opposite of "substance" and a suitable translation for this usage might be "circumstance". The modern use of the word in phrases such as "traffic accident" may have the same origin, but is otherwise rather misleading. However, the adjective "accidental" is still sometimes used to mean "circumstantial". In other words, "accident" here refers to a situation in which the normal rules do not apply and the //circumstances// demand an exception. ===== Description ===== As a general principle, following universally valid rules – be they [[wp>Statute|statutes and laws]], [[wp>Morality|ethical and moral principles]] or simply technical [[wp>Best practice|best practices]] or [[wp>Style guide|design guidelines]] – is first and foremost a positive approach. However, one should not forget that all these rules have specific //scopes// and //prerequisites// and thus also //boundaries//, outside of which they can no longer be applied in a meaningful way. In other words, there are //exceptions// under which the general rules no longer apply, or at least only with caveats. Thus, even legal provisions – such as the [[wp>Monopoly on violence|state’s monopoly on the use of force]] or [[wp>Right to property|private property rights]] – are subject to limitations, e.g. when it comes to emergencies or emergency assistance. Similar limitations also apply to other rules and standards. A principle that the author learned in connection with layout and design, and which illustrates this very well, can be formulated as follows: > //“First learn the rules,// > // then learn when to break them.”// The term "accident fallacy" thus refers specifically to the application of, and possibly insistence on, a general rule in a situation where such a "rule-breaking" would be (at least arguably) appropriate. ==== Deontological or consequentialist ethics ==== The topic discussed here has a long tradition in philosophy. It was already mentioned by Plato (see example below) and is relevant to many disciplines. However, the most striking examples probably come from //moral philosophy// ([[wp>Ethics|Ethics]]), where the //accident fallacy// marks a key point of contention between two fundamental approaches to dealing with moral questions. These approaches are known as //deontological// and //consequentialist// ethics respectively: * **[[wp>Deontology|Deontological ethics]]**((From the Ancient Greek déon [δέον] = "duty" or "obligation" and ‑logía [‑λογία] = "‑ology" or "doctrine", loosely translated: "doctrine of acting in accordance with one’s duty".)) is an umbrella term for various schools of moral philosophy that focus on adherence to fixed rules (such as, for example: "Thou shalt not kill"). * **Consequentialist ethics** (also known as "[[wp>Consequentialism|consequentialism]]"), on the other hand, describes an ethical doctrine according to which the moral value of an act depends on its //consequences//. Put simply, this could be summed up by the slogan "A good person is one who does good". These two approaches are compatible with one another in //many// situations, but contradict each other in others. Furthermore, if applied consistently, both approaches can lead to absurd, obviously "unethical" actions. The "ethically correct" application of both approaches in different situations is therefore a rich source of material for discussion within the field of [[wp>Ethics|moral philosophy]].
A good example of this conflict can be taken from the obligation //not to lie//, which, from a //deontological// perspective, is an //absolute// principle. From this point of view, it is morally //obligatory// to //always// tell the truth under all circumstances, even if this is likely to lead to negative consequences (see, for example, [[wp>Immanuel Kant|Immanuel Kant]]’s essay "[[wp>On a Supposed Right to Tell Lies from Benevolent Motives]]"). By contrast, from a //consequentialist// perspective, it is recognised that there are specific circumstances under which the general rule no longer applies, and an exception must be made. A common example for this is the need to lie in order to hide someone fleeing an act of violence from their pursuers, thereby saving their life. This was the line of argument put forward by [[wp>Benjamin Constant|Benjamin Constant]], for instance, in his essay "Des réactions politiques" (1797), which prompted Kant’s reaction mentioned above. Conversely, it is not difficult to see how such a rather [[wp>Relativism|relativist]] perspective also opens the door to abuse, namely when it can be used to justify a lie in order to gain personal advantages, or at least to avoid more tolerable disadvantages.
From a moral-philosophical perspective – and, once again, presented in somewhat simplified terms – the //accident fallacy// essentially consists in adopting a //deontological// approach in a situation where a consequentialist stance would be more appropriate((The simplification lies in ignoring the fact that established //deontological// theories do indeed recognise mechanisms for legitimate exceptions – just as //consequentialist// theories are also based on rules of conduct. The difference lies primarily in where the emphasis is placed.)). ===== Other examples ===== ==== Plato’s weapons example ==== One of the oldest descriptions of the underlying problem is the following passage from Plato’s work „[[wpde>Republic (Plato)|Republic]]“ (Politeia [Πολιτεία]): > […] Suppose that a friend when in his right mind has deposited arms with me and he asks for them when he is not in his right mind, ought I to give them back to him? No one would say that I ought or that I should be right in doing so, any more than they would say that I ought always to speak the truth to one who is in his condition.((Translated by Benjamin Jowett, original text: „[…] εἴ τις λάβοι παρὰ φίλου ἀνδρὸς σωφρονοῦντος ὅπλα, εἰ μανεὶς ἀπαιτοῖ, ὅτι οὔτε χρὴ τὰ τοιαῦτα ἀποδιδόναι, οὔτε δίκαιος ἂν εἴη ὁ ἀποδιδούς, οὐδ᾽ αὖ πρὸς τὸν οὕτως ἔχοντα πάντα ἐθέλων τἀληθῆ λέγειν.“)) >     — [[https://gutenberg.org/cache/epub/55201/pg55201-images.html#stpage331C|Platon: Republic, first book]] [331c]. Plato describes a situation in which universally applicable rules meet their limits. In this case, the main issue is the duty to keep one’s promises – specifically, the promise to return what has been borrowed. Such a duty reaches its limits when greater evil can be averted by //failing// to fulfil it. This is the case in Plato’s example when there is reason to believe that the weapon will be used to inflict harm on the owner themselves or on others – in the worst-case scenario, perhaps as the ancient equivalent of a modern-day //rampage// ([[wp>Amok syndrome|Amok]]). This intuition is not confined to antiquity: even today, no one would seriously argue that a weapon previously handed over should be returned to a person who is clearly in an acute state of mental distress (described by Plato as "not in his right mind"). ==== Pacifism vs. self-defence ==== The question of //whether//, //when// and to //whom// weapons should be given involves numerous ethical and moral considerations. Consequently, there is considerable scope for moral dilemmas that are not always easy to resolve. This applies not least to the question of the applicability of [[wp>Pacifism|pacifist]] principles. One such principle holds that conflicts should not be further fuelled by supplying the parties involved with weapons or other material essential to warfare. Otherwise, there would be a risk that a war would be unnecessarily prolonged, leading to unnecessary suffering. In principle, that is certainly a sound position. The question, however, is where such a principle reaches its limits. The current (as of July 2026) [[wp>Russo-Ukrainian war|Russian-Ukrainian war]] demonstrates how an unconditional adherence to such fundamentally positive moral principles can also lead to a dilemma: a heavily armed and militarily far superior aggressor invades a country that had neither the means nor the opportunity to prepare for such an invasion. Is it still justified, in these circumstances, to deny Ukraine the weapons and other essential military supplies it needs for [[wp>Self-defense|self-defence]]? Or is the moral duty rather to support the country that is under attack in every conceivable way? How one views this dilemma depends largely on whether one adopts the //deontological// or the //consequentialist// approach (see above), that is, whether one regards the aforementioned pacifist principle as //absolute// or judges the action by its anticipated //consequences//. ==== Rule of loyalty of subordinates ==== In hierarchical organisations, such as the police and especially the military, but also in companies or bureaucracies, there are strict chains of command in which the superiors issue instructions or orders to their subordinates. The general rule that these orders must be obeyed already results from the organic structure of these organisations, which is based on the "loyalty" of subordinates to their superiors. In many cases there is even a legal duty to obey, but where such a duty does not exist, one usually has to expect serious personal consequences if one refuses or fails to carry out orders. Nevertheless, there are of course exceptions: If such orders violate law and order, even in the most hierarchical organisation, the person receiving the orders has not only the right, but the duty to refuse them. This applies all the more when higher moral principles are involved. In other words, even if the army command orders the killing of civilians in a war, and even if this order would be "legal" according to the applicable law, this does not release individuals from their responsibility for the crime that is committed. Nevertheless, the justification that one was "only following orders" is repeatedly put forward (and just as often dismissed) when such crimes are finally tried in court. For example, this was the line of defence of [[wp>Adolf Eichmann|Adolf Eichmann]], one of the main organisers of the [[wp>The Holocaust|Holocaust]]. Unsurprisingly, it was not a successful defence either (see also: [[wp>Eichmann trial|Eichmann trial]]). Crimes against humanity or war crimes are of course extreme examples – already on a much smaller scale there are at least moral obligations to breach rules of loyalty: If, for example, one has knowledge of relevant misconduct by superiors, e.g. corruption or nepotism, this should be brought to the attention of the appropriate authorities – or if this is unsuccessful, also to the public ([[wp>Whistleblower|whistleblower]]). ==== Automated decisions / traffic cameras ==== The aspect of the "mechanical application of rules" is, not surprisingly, particularly relevant precisely when decisions are actually made by machines or automated systems. After all, these are developed specifically to operate on the basis of rules. Machines usually change their behaviour in response to new circumstances only if the altered situation has already been taken into account by their respective developers or programmers – that is, if there is a specific rule for exceptions to the norm. Consider, for example, automatic traffic cameras such as those used for red-light enforcement ([[wp>Red light camera|Red light camera]]): These cameras record incidents where a car crosses the stop line whilst the traffic lights are red. Although it is generally forbidden to drive through a red light, there are exceptions: first and foremost, the police, fire brigade and ambulances on an emergency call may disregard this rule. Furthermore, private vehicles may also cross the stop line to make way for one of these emergency vehicles. These are all exceptions to the general rule. Obviously, it would be best if the traffic cameras were "smart" enough to recognise such situations themselves and, in such cases, not to ‘flash’ the vehicles. However, this is not possible, as such systems only operate on a very simple rule: "Vehicle crosses the stop line and the traffic light is red → take a photo". These photos are then sent to the relevant enforcement agency – where a human reviewer assesses the situations and, hopefully, first sorts out all the images showing emergency response vehicles, as well as those of cars making way for them, so that the drivers do not have to worry about receiving a fine. Ultimately, the //machine// (in this case: the //traffic camera//) therefore commits an //accident fallacy//, as it does not (or can not) take into account //exceptions// to the general rule. Human intervention is required to correct this error.
Even if this isn’t a perfect solution – as we all know, people make mistakes, and there have apparently been cases where tickets were still issued for such situations (which, of course, one //can// – and //should// – appeal against!) – there is no doubt that the situation would be far worse for everyone involved if motorists were reluctant to give way to an emergency vehicle for fear of being fined.
==== Automated decisions / the “Scunthorpe problem” ==== The problem that automated decision-making processes cannot take all special cases into account becomes particularly apparent when we consider the so-called "[[wp>Scunthorpe problem|Scunthorpe problem]]": [[wp>Scunthorpe|Scunthorpe]] is a town in [[wp>Lincolnshire|Lincolnshire]], England, with a population of around 80 000. In 1996, residents of the town discovered that they were unable to register for the then-popular [[wp>AOL|AOL]] web service. It turned out that an automated obscenity filter rejected the town’s name because it contained the English word "cunt". Presumably because this was the first such incident to attract widespread attention, the town's name has come to be used as a generic term for similar incidents.((Another factor that may have contributed to this was AOL’s rather unprofessional "solution" of simply referring to the town as "Sc__o__nthorpe" instead. As might be expected, this was not particularly well received by the residents of Scunthorpe. After all, its name has a history stretching back at least a thousand years: the place name is thought to derive from the Old Norse "Skúmasþorp" ("Skuma’s village") and thus dates back to the [[wp>Viking Age|Viking Age]] (8th–11th centuries).)) The issue is always that an automated system, which is actually designed to prevent people from posting profanity for fun, is triggered even when the terms appear in a harmless context. Many of the examples of this problem are rather amusing – unless you’re the one involved, that is – therefore here a selection: * //Craig Cockburn//, a Scotsman, was unable to create an account with the online service ‘Hotmail’ because his surname contains the word "cock". * The internet domain ''shitakemushrooms.com'' (using an alternative spelling for [[wp>Shiitake|Shiitake]]) could not be registered because it contained the word "shit". * The website for the 30th edition of the American Super Bowl, known as “[[wp>Super Bowl XXX|Super Bowl XXX]]”, was blocked by internet filters. The reason: “XXX” not only stands for the Roman numeral 30, but is also used to denote pornographic content. * A Christian fundamentalist organisation set up a system on one of its web portals that automatically replaced supposedly "offensive" terms with more innocuous ones. As a result, the American sprinter [[wp>Tyson Gay|Tyson Gay]] was consistently referred to on the site as "Tyson Homosexual". * The Facebook page of the French municipality of [[wp>Bitche|Bitche]] (German: Bitsch) was automatically deleted because its name was interpreted as a variation of the English swear word "bitch". * Various Latin phrases used at universities to commend particularly outstanding achievements contain the word "cum" ("with"). The best-known example is the phrase "summa cum laude" ("with the highest honours"). However, as the English word "cum" is also a slang word for "semen", these phrases are occasionally flagged by obscenity filters. In all of the above cases, the fundamental issue is that a general rule is applied by an automated process (a word from a word list is detected → the input is blocked or replaced), without taking into account the specific circumstances of each situation. There is no simple solution to this either, except that when developing or integrating such systems, provision should always be made for human intervention. This is a lesson that would probably have spared AOL a great deal of bad press at the time. ===== See also ===== * [[logic:emergence:division|Fallacy of division]] * [[knowledge:ecological_fallacy|Ecological fallacy]] * [[generalization:hasty|Hasty generalisation]] * [[generalization:overwhelming_exception|Overwhelming exception]] ===== More information ===== * [[wp>Accident (fallacy)]] on //Wikipedia// * [[https://www.logicallyfallacious.com/logicalfallacies/Accident-Fallacy|Accident Fallacy]] on //Logically Fallacious// * [[http://fallacyfiles.org/accident.html|The Fallacy of Accident]] on //Fallacy Files//