Untergeneralisation

Untergeneralisation, or undergeneralization, is a cognitive and linguistic phenomenon in which a learner forms general rules or category boundaries that are too narrow, leading to failure to apply them to all relevant instances. It is contrasted with generalization and with overgeneralization, which applies rules too broadly.

In language development, undergeneralisation can appear when a child uses a word only in a limited context or for a single exemplar and does not extend it to other, appropriate instances. For example, a child might learn the word “dog” but use it only for one family dog and not for other dogs, or might label several different animals as “cat.” Such narrowly drawn categories reflect an insufficient generalization from observed examples.

In cognitive psychology and education, undergeneralisation describes a broader tendency to rely on overly specific rules or categories when inferring general principles from data. This can hinder flexible application of knowledge to new or varied situations, especially when encountering unfamiliar exemplars that fit the intended category but were not part of the learner’s experience.

In machine learning and artificial intelligence, undergeneralisation refers to models that generalize too little, producing narrow or incomplete predictions and failing to recognize new instances that belong to learned categories. This condition is related to, but distinct from, underfitting, which describes a model that cannot capture the underlying structure due to excessive bias or an overly simplistic hypothesis space.

Overall, untergeneralisation highlights the balance between memorized specifics and broader abstraction crucial for effective learning and communication.