Inductive fallacies involve making conclusions about a population based on observations that are not representative of the population.
Many scientific conclusions are based on inductive reasoning. Scientists seldom have access to an entire population, and therefore must perform experiments using a sample of the population. Generalizing from samples to an entire population involves inductive reasoning. Therefore, it is important to consider some of the inductive fallacies that can weaken inductive arguments. Three inductive fallacies are:
1) Unrepresentative Samples
Whenever scientists sample a population, there is the possibility that the sample does not represent the overall population in many ways. For example, women have been historically under-represented as subjects in medical studies (Kim et al., 2008). The sample of subjects for medical research is therefore un-representative of the general population (which is over 50% women; U.S. Census, 2010). There may be important sex differences in important areas such as the effectiveness (or dangers) of drugs that scientists and doctors do not know of because of historically-biased samples.
Sex/gender is far from the only ways that samples can fail to represent populations. Among the other reasons are:
A) Individual Differences: sex, age, lineage (race), socio-economic status, body weight and shape, etc. B) Environmental Differences: physical, ecological, social, economic, etc. C) Experimental Artifact: Differences in the time or methods of data collection among samples.
Ensuring that samples are representative of an overall population is challenging (and beyond the scope of our discussion). However, two common methods for making samples more representative are:
1) Balancing. If there are important variables that may affect the outcomes of an experiment, then scientists may work to "balance" samples. Balancing involves identifying important variables and ensuring that experimental groups have equal representation of different values of each variable.
2) Randomization. In many cases, scientists may not know the variables that are likely to influence an outcome. Therefore, scientists may select samples for different groups at random from a larger population. With a large enough sample size, randomization seeks to control both for variables that scientists know could influence outcomes, but also for unknown variables that are also influential.
APPLICATION: Because there are so many ways that samples can be un-representative of a population, scientists must work to ensure that experimental samples are as representative of the population as possible.
2) Hasty Generalization (Cherry Picking)
Hasty Generalization involves drawing conclusions from samples that are too small. For example:
PREMISE: Several prominent scientists falsified data. CONCLUSION: All scientists falsify data.
Clearly, the actions of a few "bad apples" cannot be used to generalize over an entire group. Hasty Generalization is a term for the common practice of coming to unreasonable conclusions based on small samples.
"Cherry Picking" is when misleading generalizations from small samples are made deliberately. For example:
PREMISE: Clarence Thomas and Barack Obama have been extremely successful. CONCLUSION: We have achieved equal opportunity in the U.S.
Cherry Picking selects small, biased samples to make an argument. There is overwhelming evidence from many areas of society that there are still substantial, systemic differences in opportunity in the United States (Hanks et al., 2018). Cherry Picking a small number of politicians is an inductive fallacy because it does not represent the bulk of the evidence.
APPLICATION: Small samples are often not representative of large populations. Simply by chance, small samples are likely to over-represent at least one variable. Moreover, deliberately "cherry picking" samples to support a conclusion does not constitute a strong inductive argument.
3) Weak Analogy
Analogies are one type of inductive reasoning (Moore and Parker, 2017). However, just as for deductive reasoning, analogies are not necessarily strong ways of making inductive arguments. For example, we could make an argument that insects are good physiological analogs for humans:
PREMISE: Humans have muscles, skeletons, nervous systems, circulatory systems, and lungs. PREMISE: Insects have muscles, skeletons, nervous systems, and circulatory systems. THEREFORE: Insects also have lungs.
Although it may seem reasonable to predict that insects have lungs, insects actually don't have lungs. Instead, insects have long tubes called "trachea" that connect the exoskeleton to internal organs and allow for gas exchange (Harrison, 2009). The analogy of insects and humans is weak (at least for gross respiratory physiology) because the areas for resemblance between insects and humans is not specific enough.
APPLICATION: Analogies can be helpful for representing and clarifying concepts or complex things. Although analogies can generate hypotheses, analogies must be quantitative and systematic to be useful scientific arguments.
Inductive reasoning can be challenging. Inductive reasoning is not truth-preserving, and fallacies can affect inductive arguments, . However, some frameworks (such as Hill's Criteria) are available to help structure inductive reasoning with hypotheses.