informal fallacies | mike bommarito

overview

informal fallacies are errors in reasoning that stem from problems with content, context, or assumptions rather than logical structure. unlike formal fallacies, these errors often involve misleading information, irrelevant details, or false presuppositions that undermine the argument’s strength.

understanding informal fallacies is crucial for evaluating everyday arguments, media claims, political rhetoric, and any reasoning that goes beyond pure formal logic.

characteristics

content-dependent errors

informal fallacies involve problems with what the argument says, not how it’s structured:

formally valid but informally fallacious:
if experts recommend X, then X is good          (premise)
this tobacco company expert recommends smoking  (premise)
therefore smoking is good                       (conclusion)

logical form: valid (modus ponens)
content problem: irrelevant authority (tobacco company bias)

context sensitivity

the same content might be fallacious in one context but reasonable in another:

context 1 (medical): "dr. smith recommends this treatment" → reasonable appeal to authority
context 2 (legal): "dr. smith says the defendant is guilty" → irrelevant authority

pragmatic factors

informal fallacies often involve:

psychological manipulation: exploiting cognitive biases
rhetorical tricks: persuasion through irrelevant means
conversational context: violating discourse norms
background assumptions: relying on questionable presuppositions

major categories

fallacies of relevance

premises that don’t actually support the conclusion:

ad hominem

attacking the person making the argument instead of the argument itself

sarah argues for higher teacher salaries
but sarah was a terrible student herself
therefore her argument is wrong

red herring

introducing irrelevant information to distract from the real issue

discussion: should we increase education funding?
response: "what about all the corruption in other government programs!"

appeal to emotion

using emotions instead of relevant evidence

we must ban this chemical because think of the children!
(without evidence the chemical actually harms children)

fallacies of presumption

arguments that assume what they’re trying to prove or make unjustified assumptions:

begging the question (circular reasoning)

using the conclusion as support for itself

the bible is true because it says so in the bible

false dilemma

presenting only two options when more exist

"you're either with us or against us"
(ignoring neutral, partial support, or alternative positions)

loaded question

asking a question that contains controversial assumptions

"when did you stop cheating on your taxes?"
(assumes the person was cheating)

fallacies of ambiguity

arguments that exploit unclear or shifting meanings:

equivocation

using the same word with different meanings

premise: laws of nature are universal (scientific laws)
premise: laws require a lawmaker (legal laws)
conclusion: therefore nature has a lawmaker

problem: "law" means different things in each premise

amphiboly

ambiguous grammar creates misleading meaning

"the man hit the boy with the stick"
unclear: who had the stick?

composition/division

what’s true of parts/whole is true of whole/parts

composition: each player is excellent, so the team is excellent
division: the team is excellent, so each player is excellent

common informal fallacies

appeal to authority

citing irrelevant, false, or inappropriate authorities:

fallacious forms:

irrelevant expertise: “einstein believed in god, so atheism is wrong”
false authority: “my neighbor says climate change is fake”
popular authority: “most celebrities support this candidate”

legitimate forms:

relevant expertise: “oncologists recommend this cancer treatment”
consensus of experts: “97% of climate scientists agree on global warming”

slippery slope

claiming one event will inevitably lead to a chain of bad consequences:

fallacious: "if we allow gay marriage, next people will marry animals"
reasonable: "if we reduce safety inspections, accident rates will likely increase"

detection criteria:

are the causal connections actually supported?
is the slope actually slippery or can we stop at reasonable points?

hasty generalization

drawing broad conclusions from insufficient evidence:

fallacious: "i met two rude french people, so all french people are rude"
reasonable: "in this sample of 10,000 voters, 55% support candidate X"

false analogy

comparing unlike things and drawing invalid conclusions:

fallacious: "employees are like family, so we don't need formal contracts"
reasonable: "like biological evolution, ideas evolve through variation and selection"

detection strategies

relevance testing

ask: “do the premises actually relate to the conclusion?“

argument: "john's proposal for traffic reform won't work because he's divorced"
relevance test: does marital status relate to traffic expertise? → no → ad hominem

assumption analysis

identify hidden assumptions:

argument: "we must choose between economic growth and environmental protection"
hidden assumption: these are mutually exclusive → false dilemma

context evaluation

consider the domain and situation:

"trust me, i'm a doctor"
medical advice: potentially relevant
stock market advice: irrelevant authority

emotional vs rational content

separate emotional appeals from logical evidence:

"this policy will destroy our way of life!"
question: what specific harms are predicted? what evidence supports these predictions?

automated detection challenges

natural language processing

informal fallacies are harder to detect automatically than formal ones:

# pseudocode for informal fallacy detection
def detect_informal_fallacies(text):
    # challenges:
    # - semantic understanding required
    # - context-dependent
    # - cultural/domain knowledge needed

    fallacies = []

    if contains_personal_attacks(text):
        fallacies.append("potential_ad_hominem")

    if identifies_false_dichotomy(text):
        fallacies.append("potential_false_dilemma")

    # many more heuristics needed...
    return fallacies

machine learning approaches

# training data for fallacy classification
def train_fallacy_classifier():
    training_data = [
        ("john's idea is bad because he's ugly", "ad_hominem"),
        ("either we ban cars or accept pollution", "false_dilemma"),
        ("everyone thinks X, so X must be true", "appeal_to_popularity"),
        # thousands more examples needed...
    ]

    # challenges:
    # - need large labeled datasets
    # - context understanding crucial
    # - cultural/domain variations

    return train_classifier(training_data)

knowledge-based detection

# rule-based approach with domain knowledge
def detect_with_rules(argument):
    if contains_authority_claim(argument):
        authority = extract_authority(argument)
        topic = extract_topic(argument)

        if not is_relevant_expert(authority, topic):
            return "irrelevant_authority"

    if contains_causal_chain(argument):
        chain = extract_causal_chain(argument)

        if not well_supported_chain(chain):
            return "slippery_slope"

    return "no_fallacy_detected"

real-world examples

political discourse

ad hominem: "don't trust senator X's tax proposal - he's been divorced three times"
straw man: "my opponent wants to defund the police" (when they proposed budget reallocation)
false dilemma: "we must choose between security and privacy" (ignoring balanced approaches)

advertising

appeal to emotion: "don't let your family suffer - buy our insurance"
bandwagon: "millions of people can't be wrong - join now!"
false authority: "four out of five doctors recommend..." (doctors of what? which ones?)

media analysis

hasty generalization: "crime is rising everywhere" (based on selective reporting)
red herring: discussing politician's personal life instead of policy positions
loaded question: "when will the administration stop ignoring the crisis?"

scientific discussions

appeal to authority: "famous physicist believes in psychic powers"
false analogy: "if quantum mechanics is weird, anything is possible"
composition: "this study shows X works for these patients, so it works for everyone"

contextual considerations

cultural variations

fallacies can be culture-dependent:

appeal to authority: varies by how much cultures respect hierarchy
ad hominem: some cultures consider character more relevant to arguments
false dilemma: cultures with different conflict resolution styles

domain-specific patterns

different fields have characteristic fallacies:

legal: irrelevant evidence, emotional manipulation of juries medical: appeal to nature fallacy, post hoc ergo propter hoc business: false dilemma in decision-making, hasty generalization from limited data politics: ad hominem attacks, straw man misrepresentations

technological context

social media and online discourse create new patterns:

echo chambers: amplify confirmation bias
viral misinformation: spread hasty generalizations
character limits: encourage oversimplification (false dilemmas)
anonymous posting: enable ad hominem without accountability

avoidance strategies

constructing arguments

focus on relevant evidence
- ensure premises actually support conclusions
- avoid personal attacks or irrelevant details
acknowledge complexity
- consider multiple options, not just two
- recognize nuanced positions
cite appropriate authorities
- use experts relevant to the topic
- acknowledge limitations of authority
avoid emotional manipulation
- separate emotional appeals from logical arguments
- use emotions appropriately (e.g., moral arguments)
define terms clearly
- avoid equivocation
- ensure consistent use of concepts

evaluating arguments

separate content from form
- check logical structure first
- then examine content quality
identify the real issue
- ignore red herrings and distractions
- focus on the central claim
check assumptions
- what unstated beliefs are required?
- are these assumptions reasonable?
consider alternatives
- are there other options besides those presented?
- what middle ground exists?
evaluate evidence quality
- is the evidence relevant and adequate?
- are sources credible and appropriate?

teaching informal fallacies

progressive approach

start with clear examples
- use obvious cases first
- gradually introduce subtlety
practice identification
- analyze real arguments from media, politics
- distinguish fallacies from legitimate reasoning
understand context
- same content can be fallacious or reasonable
- cultural and domain variations matter
application exercises
- rewrite fallacious arguments to be stronger
- construct counterarguments without fallacies

common student challenges

over-identification: seeing fallacies everywhere
context blindness: ignoring when appeals are legitimate
false equivalence: thinking all arguments with fallacies are equally bad
missing subtlety: only recognizing obvious cases

relationship to cognitive biases

many informal fallacies exploit psychological biases:

fallacy	exploits bias
ad hominem	halo effect, in-group bias
appeal to emotion	affect heuristic
bandwagon	social proof, conformity
confirmation bias	cherry-picking evidence
availability heuristic	hasty generalization

understanding these connections helps both identify fallacies and resist their influence.

computational applications

argument mining

# extract and evaluate arguments from text
def mine_arguments(text):
    arguments = extract_argument_structures(text)

    for arg in arguments:
        formal_validity = check_logical_structure(arg)
        informal_issues = detect_informal_fallacies(arg)

        yield {
            'argument': arg,
            'formally_valid': formal_validity,
            'informal_fallacies': informal_issues,
            'overall_strength': evaluate_strength(arg)
        }

fact-checking systems

# integrate fallacy detection in fact-checking
def enhanced_fact_check(claim, supporting_arguments):
    factual_accuracy = check_facts(claim)
    logical_structure = analyze_logic(supporting_arguments)
    fallacy_analysis = detect_fallacies(supporting_arguments)

    return {
        'factual': factual_accuracy,
        'logical': logical_structure,
        'fallacies': fallacy_analysis,
        'overall_credibility': compute_credibility_score()
    }

dialogue systems

# avoid fallacies in AI-generated arguments
def generate_argument(position, evidence):
    # ensure relevance
    relevant_evidence = filter_relevant(evidence, position)

    # avoid common fallacies
    argument = construct_argument(position, relevant_evidence)

    if contains_fallacies(argument):
        argument = revise_argument(argument)

    return argument