Statistics & Coding
This immutable Python data type is an ordered sequence of characters, like "Hello, World!".
String
This type of sampling bias occurs when a sample is collected in a way that some members of the intended population are less likely to be included than others (e.g., online surveys excluding people without internet).
Selection Bias
In a facial recognition system set to identify suspects, this is the undesirable outcome when a criminal is scanned but the system fails to flag them.
False Negative
This common fairness measure ensures the True Positive Rate is equal across different demographic groups.
Equal Opportunity
This lens judges an action as ethical if it produces the greatest good for the greatest number of people.
Utilitarianism/Consequentialism
The classic phrase warning that just because two things trend together doesn't mean one causes the other.
Correlation does not imply causation
This citation style, required for many social science courses, has specific rules for citing sources to avoid this form of bias.
APA format
This is the term for a correct identification, such as when a face is correctly matched to its database entry.
True Positive
If a hiring algorithm approves a higher percentage of men than women who are equally qualified, it violates this measure of fairness, which focuses on matching selection rates.
Demographic Parity
You argue that lying is always wrong, even to save someone's feelings, because it violates a universal moral rule. You are using this lens.
Deontology/Duty Based Lens
The practice of manipulating data analysis or the research process until a statistically significant (p < 0.05) result is found.
p-hacking
This type of sampling gives every member of the population an equal and known chance of being selected, like drawing names from a hat.
Simple Random Sampling
This 1970s committee decided which patients with kidney failure would receive access to the then-scarce resource of dialysis machines.
The God Squad
This is the probability that someone who tests positive (or is flagged by a model) actually has the condition or is a correct match.
Precision or PPV
If you need to look up the official principles on avoiding harm, being honest, and respecting privacy in computing, you would consult this document.
The ACM Code
Unlike linear regression used for continuous outcomes, this type of regression is used when your dependent variable is binary (e.g., 0 or 1).
Logistic/Logit regression
This is the error from a model that is too simple, like using an average to predict everything.
Bias
This company's credit card algorithm came under fire for offering significantly higher credit limits to men than to women with similar financial profiles.
Apple Card/Goldman Sachs
You cannot achieve both Demographic Parity and this other fairness measure (which requires similar error rates across groups) if the base rates of the outcome differ between groups.
Predictive Parity or or Equal False Positive Rates
If you oppose a new social media feature because it could erode community trust and human connection, you are primarily arguing from this lens, which focuses on the character of society.
Virtue Lens
In Python, this library, commonly imported as pd, is the fundamental tool for working with structured data frames.
Pandas
A key strategy to combat high variance is to gather more of this, which helps the model learn the true signal instead of the noise.
Training data
This retailer famously used purchasing data to predict customer pregnancy and send targeted coupons, sometimes before the family had announced it.
Target
In this 2014 case, an algorithm used for criminal sentencing was found to be disproportionately flagging Black defendants as high-risk.
COMPAS
This trap occurs when you try to solve a problem by focusing only on the technical system, while ignoring the social context that shapes its use and impact.
Framing Trap