How to Analyse Arguments From Analogy

[important]”Analysing Arguments” is an ongoing series which analyses arguments found in daily life. Some good background material for this is Coursera’s enormously popular course Think Again: How to Reason and Argue, and the book Understanding Arguments: An Introduction to Informal Logic. You might also find the primer How to Argue Online useful. Other installments of the series are listed in the analysing arguments tag here and also on[/important]

An argument from analogy is similar to what we simply call an analogy, but is different in that it’s an argument, whereas an analogy is usually just a stating of a similarity. “Your driving is like Rahul Dravid’s batting” – that’s an analogy (and a compliment I once received!). It’s simply saying A is like B. But an argument from analogy – henceforth referred to as an “AFA” – is an inductive argument, which states the existence of a further similarity as its conclusion. It takes the following form:

1. Object A has property P (and possibly Q, R…).

2. Object B also has property P (and Q, R…).

3. Object B has property X.


4. Object A also has property X. (From 1-3.)

Here’s an example which some of us might have experienced. Say I’m up for a promotion at work, and I think I’ll be promoted, but I’m not. And then I find out that my colleague has been promoted. So I go to my boss and I argue for my promotion by saying:

“You promoted <colleague>, why didn’t you promote me?”

I’ll try to reconstruct this AFA in the above form. First, what are the two objects being compared? Easy enough – me and my colleague. I.e.,

A = me

B = colleague

Next, what are the common properties P, Q, R and so on? It’s implicit that there must be some similarities between my colleague and I, so let’s say we both joined around the same time, and have similar experience levels:

P = 5 years of relevant experience

Q = joined the company in 2009

Finally, X is the property of being promoted:

X = got promoted

So as you can now see, the conclusion “A also has property X” is “I should also be promoted”.

When is an Argument From Analogy Strong?

An AFA is stronger when it has the following attributes:

  • Many relevant similarities: the similarities P, Q, R… are relevant to X and many in number. The similarities I noted above are certainly relevant to the issue of promotion. But other similarities might not be relevant – e.g. if my colleague and I both have degrees in philosophy, but philosophy isn’t relevant to our job, then that similarity isn’t relevant to our promotion. But if my colleague and I both hit a particular sales target in the last quarter, and we both won a particular award… the more relevant similarities there are, the stronger the argument becomes.
  • Fewer relevant dissimilarities: there are fewer relevant dissimilarities between A and B. What if it turns out that my colleague passed a well-regarded industrial certification, and I hadn’t? Or received previously-unheard of praise from the customers? Or won major new business? Or hired and coached a brilliant team? If I have done none of these things, then these differences are relevant, and they would weaken my argument.
  • Diverse objects: there are other objects C, D, E… which also have similarities P, Q, R… and X. If I can identify three or four other colleagues who also share those similarities and got promoted, then my case for promotion becomes stronger.
  • Weaker conclusion: If instead of saying “You definitely should have promoted me”, I say “You probably should have promoted me”, the argument becomes stronger. Granted, in this particular example it wouldn’t make much sense, since promotion is a yes-or-no state. But in general, the principle holds – a weaker conclusion has more support from the premises of an AFA.

Here’s a real-life AFA from a few days ago – in a Wall Street Journal interview, the CEO of American financial services firm AIG said this while responding to criticism of AIG executives receiving bonuses despite the company being in bad shape:

The uproar over bonuses was intended to stir public anger, to get everybody out there with their pitch forks and their hangman nooses, and all that–sort of like what we did in the Deep South [decades ago]. And I think it was just as bad and just as wrong.

He later apologised (kind of). Most of us can instinctively make out what’s wrong with this argument, but it helps to break it down into the above form, to see just why it’s a weak argument:

A = executives of AIG who received bonuses

B = African-Americans in slavery/civil rights era

P = demonised by media and public opinion

X = ought to be left in peace.

The argument appears to point out an additional similarity by use of the phrases pitchforks and hangman’s nooses, but for one object (A) these phrases are rhetorical, but for the second (B) they are literal (an equivocation fallacy perhaps?). So we’re left with that one similarity P, of questionable relevance, and of course there are a host of relevant dissimilarities between A and B. As a result, this is a very weak argument from analogy.

A more detailed and academic look at AFAs can be found at the Stanford Encyclopedia of Philosophy. I plan to continue with the same subject next time, where I’ll look at arguments from analogy which routinely show up in the act of victim-blaming.



  1. iplon says

    I really like your formulating of this argument and how to qualify it’s strength. It reminds me of a lot of neat stuff that I have to do with machine learning. A “simple” question somebody might ask is “what qualities are best for identifying somebody that deserves to be promoted/hired”. We’d take a lot of recent data about people, with people grouped into classes like “Promoted”, “Demoted”, “No Change”, “Hired”, “Fired”, and we’d run a lot of complicated analysis on it and eventually train a program that can actually, with a certain (roughly known) accuracy, take the same data about somebody else and predict what class they would go into.

    You’ve done something that I think could be fascinating and twisted this into a new usage. If we can create and train a program to analyze what factors are most useful in determining what category someone will go into, you have now pointed out to me that we should also be able to ask, “Why was this person put in the wrong class?” This is especially important to the person who, while all variable seem to point out they should have been promoted, were actually kept the same (or worse: demoted or fired).

    This could also be very important for social justice issues. If you find out that race or gender is actually is a good predictor (that is, strengthens the accuracy of predictions by the program), you’ve just highlighted a major problem.

  2. sezit says

    I have long felt that the best comebacks to insults or callous comments are AFA that the other person has to think through. Here’s 3 real-life examples:
    1. When a (black male) supervisor stated proudly: “there’s my girls, working on the problem!”, I asked him how he would like it if I called him “my boy” (which in the US has overtones of slavery)
    2. A manager, upon learning that I had a 15 yo cat, said “why don’t you put that thing to sleep and give another animal a chance”. My reply was “I heard your parents were getting old.” (I didn’t even know if they were alive, but I was that pissed off. My cat lived another 5 years.)
    3. When Flo Kennedy (a huge feminist activist) was heckled at speeches by men calling out “You’re a Lesbian!”, her response was “And you are the alternative?”
    The best part is that these comebacks can be said sweetly or deadpan, so that the insulter is working out the meaning while you have moved on.

Leave a Reply

Your email address will not be published. Required fields are marked *