SUMMARY - Calling Bullshit - Jevin D. West & Carl T. Bergstrom



Here is a summary of the key points:

  • Misinformation and disinformation spread easily on social media due to the incentives for rapid sharing without fact-checking. This has real-world consequences in some cases.

  • Even educated people can be misled by false stories. Political propaganda takes advantage of social media to potentially influence large numbers of people.

  • "Firehose" tactics of spreading high volumes of contradictory information overwhelm critical thinking. Governments also use social media for surveillance and opinion manipulation.

  • Fake news stories, fake accounts/bots, and deepfake technology undermine trust in information on online platforms by complicating efforts to verify the truth. However, cross-checking multiple independent sources remains a viable strategy.

Overall, the chapter discusses how social media'sarchitecture and dynamics have enabled widespread spread of misinformation, with serious potential implications if left unchecked. Verifying information online remains challenging.

Here is a summary of the key points:

  • Numbers and statistics are widely used to analyze and describe the world, but they must be interpreted carefully to avoid misinformation.

  • Many numerical claims are based on samples rather than full counts or measurements due to feasibility. This introduces potential errors like sampling bias or measurement error.

  • Context around how data was collected and potential limitations is important for accurate understanding. Numbers alone can be manipulated or misconstrued.

  • Samples must be truly representative of the overall population to generalize findings. Biases in sampling methodology can skew results.

  • Even studies with large sample sizes require evaluating how participants were selected and if certain groups are under or over-represented.

  • Context is needed around numbers to avoid taking them at face value or being misled. While quantitative, numbers still require qualitative judgment around their validity and limitations.

The key point is that numbers must be interpreted carefully with awareness of methodological limitations and context, rather than assuming they provide a comprehensive or unbiased perspective on their own. Proper understanding requires evaluating both quantitative findings and qualitative assessment of how data was gathered.

Here is a summary of the key points about Berkson's paradox:

  • Berkson's paradox occurs when two variables appear correlated due to a pre-selection bias, even though they are statistically independent in the overall population.

  • It helps explain the perception that attractive people tend to be jerks. Attractiveness and niceness are likely uncorrelated when considering all people.

  • But when choosing potential dating partners, people select for both high attractiveness and high niceness. This creates a selection bias.

  • Among the subset of people an individual would consider dating, there is now an artificial negative correlation between attractiveness and niceness. More attractive options seem nicer on average.

  • This is because very attractive but unnice people are excluded from consideration due to the minimum niceness threshold. Their exclusion makes attractiveness seem inversely related to niceness among options under consideration.

  • But in the overall population of all people, no real correlation exists between these traits. The perceived relationship is an artifact of the selection process, not the variables themselves.

So in summary, Berkson's paradox describes how pre-selecting individuals based on multiple criteria can induce artificial correlations between variables that are truly independent.

Here is a summary of the key points from the passage:

  • Line graphs are generally better than bar graphs for showing trends and patterns over time. However, bar graphs can be useful for comparing categories at a single point in time.

  • When using bar graphs to compare values over time, it is important to ensure the y-axis starts at zero so the lengths of the bars accurately represent the magnitudes being compared. Starting above zero can make differences appear larger.

  • 3D graphs should generally be avoided as they can distort visual perception and misrepresent relationships. Perspective means bars further from the viewer will appear artificially shorter even if they represent equal values.

  • When interpreting any graph, it's important to carefully examine things like axis scales, units of measurement, categories/groups being compared, and whether relevant context is provided or omitted. Small design choices can inadvertently or intentionally mislead.

  • Simple, clear visualizations that accurately represent the raw data in an unbiased way are ideal. Graphical elements should enhance understanding, not distort perception or tell a story not fully supported by the numbers.

So in summary, choosing the right graph type, ensuring accurate axes, and providing full transparency are keys to creating data visualizations that inform rather than mislead the viewer.

Here is a summary of the key points about p-values:

  • P-values indicate the probability of obtaining results at least as extreme as what was observed if the null hypothesis is true, not the probability that the null hypothesis is false or the alternative is true.

  • Mistaking a p-value for the probability that the null hypothesis is false is called the "prosecutor's fallacy" - incorrectly implying strong evidence against the null like a prosecutor implying guilt from a fingerprint match.

  • Directly calculating the probability of the alternative hypothesis is difficult as it depends on prior beliefs, which often differ. P-values only consider the null.

  • P-values are often inaccurately presented outside scientific circles, like implying a 99% certainty of a discovery from a 1% p-value.

  • Proper interpretation is important to avoid "bullshit" claims not supported by evidence and for scientific reasoning. Low p-values do not on their own prove strong evidence against the null.

So in summary, p-values are frequently misunderstood or misrepresented, and their true meaning in evaluating evidence against the null hypothesis needs to be clear to draw valid scientific conclusions.

Here is a summary of the key points made in the passage:

  • Headlines, lists and comparisons can sometimes be misleading if they ignore important context or definitions.

  • Two examples are given: a headline comparing germs on airport trays and toilet seats without noting different types of germs were measured, and a list of "most dangerous cities" that may be skewed by how city boundaries are drawn.

  • For cities, tighter boundaries that only encapsulate the urban core could result in higher reported crime rates than looser boundaries including suburbs.

  • This hypothesis could be tested by comparing a city's size to its surrounding metropolitan area - expecting tighter boundaries to correlate with higher average crime.

  • Understanding how metrics, comparisons and datasets are constructed is important for properly interpreting lists, rankings and analyses that aim to compare or classify complex phenomena. Context matters to avoid unfair or misleading impressions.

In summary, the passage cautions against taking headlines, lists and comparisons at face value, arguing their meaning depends greatly on contextual details about methodology, definitions and data collection that are often overlooked.

Here is a summary of the key points about effectively calling out bullshit:

  • Thoroughly research facts yourself before making assertions in a rebuttal. Double and triple check sources to ensure accuracy.

  • Approach rebuttals charitably instead of assuming malice or incompetence on the part of the original speaker. Consider the possibility of honest mistakes.

  • Admit errors gracefully if you make a mistake yourself instead of doubling down. Owning mistakes maintains credibility.

  • Present rebuttals clearly and understandably instead of through an overload of facts or jargon. Effective refutation takes work to communicate persuasively.

  • Focus criticisms on the argument, not the person making it. Attack the ideas, not the individual.

  • Ensure corrections are pertinent to the actual discussion, not irrelevant "well actually" points.

  • Consider the audience and whether confronting a claim will convince anyone or simply be confrontational. Pick battles wisely.

The overall message is that effectively rebutting incorrect information requires thoroughness, charity, honesty, clarity, relevance and judgment - not just pointing out errors, but advancing truthful understanding.

Here is a summary of the key points made in the tweet:

  • Benjamin Wittes poses a hypothetical scenario about finding unattended candy on the street to make an analogy.

  • The scenario prompts consideration of whether one would eat the candy themselves, or give it to others like children or friends, without knowing its origin or safety.

  • This analogy relates to how people should thoughtfully evaluate information found online, just as they would consider an unknown food's safety before consumption.

  • The tweet encourages skeptical and prudent assessment of claims, rumors, and statements spread via platforms like Twitter, rather than automatically trusting or spreading unverified information just because it is accessible.

  • It prompts readers to think critically about what kind of "information candy" they would thoughtlessly put in their mouths or share further without verification, as they would avoid uncertain food left unattended on the street.

In summary, the tweet uses a relatable food scenario to invoke careful consideration of how credulously one consumes or spreads digital information without verification, in the same way one would avoid questionable physical items without knowing their provenance or composition.

Did you find this article valuable?

Support Literary Insights by becoming a sponsor. Any amount is appreciated!