We know that arguing from one anecdote is fallacious.

What about multiple anecdotes describing the same phenomenon? For example, using customer reviews to buy a product. Would this still be considered fallacious?

Edit/clarification: If a lot of people have a similar experience and say something is true, then would it be true?

In the case of customer reviews, it would just be people who decided to leave a review. A customer who had an average experience or did not want to spend 5 minutes typing something out might not be inclined to leave a review. If a lot of customers leave a positive review, would that mean that the product is objectively reliable? What would happen if another group of people left the same amount of negative reviews as positive reviews?

A potential counter example to this would be the existence of God. Some people who have a near death experience claim that God came to them. Does that support the existence of God?

In these various contexts, are multiple anecdotes still logically unsound?

  • I suppose customer reviews are a form of appeal to authority. May 25, 2022 at 4:57
  • 2
    You might be using the wrong context here. Reasoning is fallacious if the premises do not guarantee the conclusion must also be true within deductive reasoning. Inductive reasoning allows there to be possible false conclusions from the premises. Deductive reasoning does not allow probability. Fallacious reasoning could be the rejection of a conclusion based on a bias about the topic. A common tactic is to change the focus to something positive like a review so one would accept the conclusion given by someone. Either way we are accepting or rejecting a conclusion. The Why is very important.
    – Logikal
    May 25, 2022 at 12:24
  • 1
    You're in a little town in a foreign country. You're hungry. There are two sidewalk food vendors side by side. One has no customers. The other has a line. Where do you eat? You get at the end of the line, right? That's not an "appeal to authority," that's the common sense that you can't possibly know all the underlying factors, but you can leverage what the locals know. You'd be a fool to be the only one eating at the place where none of the locals will eat. And absent the problem of fake reviews, online reviews are the best way to know what to buy. You can't personally investigate everything.
    – user4894
    May 25, 2022 at 15:05
  • Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking.
    – Community Bot
    May 25, 2022 at 17:00
  • 1
    It depends on the origin of the anecdotes. If they are more or less independent, as, presumably, customer reviews are, but agree with each other Bayesian updating will increase the initial confidence in the underlying fact steadily. The opposite situation is when they all spring from the same source, as happens with historical memes, often unbeknownst to the reader or listener, and there is no independence at all.
    – Conifold
    May 25, 2022 at 18:01

3 Answers 3


Short Answer

Are multiple anecdotes still logically unsound?

Not necessarily.

Long Answer

Honestly, it depends on the context of logic. Anecdotal evidence isn't a cut and dry matter. From the article:

For instance, in medicine, published anecdotal evidence by a trained observer (a doctor) is called a case report, and is subjected to formal peer review.2 Although such evidence is not seen as conclusive, researchers may sometimes regard it as an invitation to more rigorous scientific study of the phenomenon in question.

In science under the logic of induction, the nature of the anecdote would determine its degree of strength. Science is greatly concerned with observations, and in science, a single anecdote might form the backbone of a counterexample or a "major piece of the puzzle". A medical doctor who provides testimony may help confirm or disconfirm a conclusion during differential diagnosis.

For instance, in a court of law, expert witness evidence might include testimony that is anecdotal, and that might be enough to make or break an argument in the eyes of the trier of fact. Again:

Expert witnesses are present in litigation to explain complicated scientific issues, not to influence the jury or judge with fervor. The main responsibilities of expert witnesses are to evaluate potential problems, defects, deficiencies, or errors only when able to fully appreciate a process or system.

And not all use of anecdote is invalid. Many simple minds reason like this. Someone mentioned an authority in their argument. Appeal to authority! He's a manager. Appeal to Force! There's an anecdote in the argument. Argument from anecdote! What makes the anecdote fallacious is it's use as a type of hasty generalization; that is, the argument seems to hinge on the implicit assumption that the anecdote asserts itself as infallibilism. For instance:

Johnny Badlogic: I met a brown bear once, and it was obvious that brown bears are gentle unlike their polar cousins since the bear was kind to me.
Billy Reasonable: You understand that the bear in question was trained for movies and raised his whole life among people, right?
Johnny Badlogic: A bear's a bear's a bear!

Contrast that to this:

Peter the Scientist: I've been studying otters all of my life, and I've only seen an otter in the wild attack a person once. Generally, they flee when people come around. I think the safe conclusion is that otters do not have a disposition (SEP) to attacking people.
Mary Judgesthingsforaliving And your conclusion therefore is that the witness's claim that he was chased down the street by 20 otters necessitating his breaking into the gun shop seems highly improbable.
Peter the Scientist: Well, while attacks by otters do happen on occasion, that claim seems to presume a certain level of determination that has never been reported of otters. And certainly the claim they followed upstairs and trapped him on the roof shortly before the police arrived and then ran off without a trace seems... farfetched?

Lastly, remember that as anecdotes are testimony, multiple anecdotes often form the back ground of scientific research, particularly under the heading of self-reporting. Longitudinal studies conducted to study topics, such as happiness in positive psychology, are one of the backbones of psychological study, for instance. One anecdote in an argument about UFOs is one matter. But 800 people contributing anecdotes over 40 years of their life is another. There's significant strength of epistemic validity in polling particularly when well-devised protocols are observed. Where is the line between a single anecdote and a longitudinal study? Well, that's the sorites paradox at play.

Yes, a million anecdotes don't necessarily prove a truth. Remember, truth is a tricky subject. For instance, self-reporting might be flawed, as the Dunning-Kruger effect demonstrates. And such arguments are likely appeal to popularity:

A majority of the people in the world believe in the supernatural because they have an anecdote. That many people can't be wrong. Therefore, there is a God of some sort.

Consider the same argument with a different topic:

A majority of the people in the world believe the earth is flat based on their lack of experience of it being anything but flat generally. That many people can't be wrong. Therefore, the earth is clearly not an oblate spheroid.

So, a knee-jerk thinker (and we have quite a few of them on this site) will "fallacy monger" when an anecdote enters play and blurt out "Nuh-uh! Anecdote!", but the sophisticated thinkers realizes that anecdotes are testimony, which is a legitimate epistemic source of knowledge, and quality, quantity, and context matter.

  • 1
    Thanks for the answer. Polling vs self-reporting - Anecdotes would have to be collected in a controlled manner to reduce bias.
    – DdogBoss
    May 26, 2022 at 16:37

It depends on your hypothesis.

For example, if your hypothesis is that black swans do not exist, a single anecdote of "I saw a black swan in Australia" is evidence against the hypothesis. It is not proof, as the person may be lying or mistaken. In this instance, a lot of people saying the same thing does indeed strengthen the evidence.

When it comes to online reviews the value of any observation will depend on your beliefs/model. For example, if one held that the proportion of good to bad reviews was proportional to the quality of the product, then an observation that product A had a higher ratio of good reviews to bad than product B would be evidence that product A is better than product B. If your model was that good products have real online reviews that are 90% good and 10% bad and no fake reviews, and bad products have 10% real reviews that are all bad and 90% fake reviews that are all good then an observation that product A had a higher ratio of good reviews to bad than product B would not be evidence at all.

  • Ok. I think we are making progress. Multiple experiences would be useful depending on the model, hypothesis, and who is saying it. I see how it could be an appeal to authority. If someone left a review of a product, that would imply that they are briefly a domain expert but there would still be unknowns on how that one person used the product. The model would have to take into account the authority of the population that left reviews. @Logikal was correct when they were saying we are dealing with a conclusion. What happens if you use a different model from others?
    – DdogBoss
    May 26, 2022 at 14:48

Anecdotes are a valid approach to logic. Inductive logic is nothing if not a collection of anecdotes (called evidence) which leads to a conclusion. What the idea that arguing from an anecdote is fallacious comes from is that this is not a valid form of deductive reasoning.

Truth be told, in many situations, inductive reasoning is acceptable, so the logic is sound. It all depends on the context and what sort of argument is being made. Sometimes people will only accept a deductive argument, and in those circumstances relying on anecdotes will forever be unsound.

As Conifold wrote in the comments, there is an entire school of probability theory called Baysean Statistics, which centers around updating "beliefs" about random variables given anecdotes. It is very specific in what it does, in order to remain fully in the domain of probability theory. I'm a real fan of Cassie Kozyrkov's Are you Baysian or Frequentist as a quick and intuitive foray into that world.

Even if you do go down the path of Baysean statistics, you eventually have to make a decision. Quite often we rely on abduction to do this. Abduction lets us say "It is most likely that this product is good, thus I assert that it is good." This is the last step in your produce purchase argument. Once again, this is decidedly not valid deductive logic. It's abductive. But it is hard to find a philosopher who will argue against pragmatic value of this approach for most arguments. Only with a handful of particularly touchy subjects (such as the existence of God), or strict topics such as mathematics, might someone refuse to accept your abductive logic when a deductive logic would have been accepted.

  • Thanks for the answer. This was helpful.
    – DdogBoss
    May 26, 2022 at 16:43

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .