Machine Learning


Strategies used by computers to surpass their designers’ conceptual competencies

Principal Metaphors

  • Knowledge is … task-based competence
  • Knowing is … operating
  • Learner is … a mechanical processor
  • Learning is … generating inferences that surpass inputs
  • Teaching is … programming




Machine Learning is a branch of Artificial Intelligence. It encompasses all efforts and strategies to structure and program machines to improve their performances on specific tasks. In these applications, “learning” is usually defined in operational rather than cognitive terms – that is, in terms of measurable improvements in performance that are due to changes in the machine’s structure, programs, or data. Many strategies are under development, employing different computer architectures and a range of logics. Prominent approaches include genetic algorithms (mimicking the process of natural selection), Bayesian networks (based on probabilistic modeling of a situation), support vector machines (which apply sorting techniques to classify phenomena), and artificial neural networks (Deep Learning; modeling brain structure to perform complex recognitions). Associated discourses include:
  • Active Learning – In the context of Machine Learning, Active Learning refers to strategies aimed at maximizing performance while minimizing trials/samples. (Note: Should not be confused or conflated with Active Learning.)
  • Actor–Critic Model ­– a two-component, iterative model of learning that combines the activities of an actor (which establishes associations between stimuli and responses) and a critic (which assesses the value of stimuli)
  • Algorithmic Learning Theory (Algorithmic Inductive Inference) (Mark Gold, 1960s) – a gradual strategy for “teaching” a program to learn the grammar of a language by requiring it to infer rules as it assesses the grammatical correctness of a sequence of statements
  • Apprenticeship Learning (Learning from Demonstration) – Invoking the notion of Apprenticeship developed within Situated Learning (which involves working with an expert), Apprenticeship Learning is a form of Supervised Learning (see above) in which the dataset comprises a carefully selected set of efficient task executions.
  • Automated Learning – 1. a synonym for Technology-Mediated Individual Learning. 2. a synonym for Machine Learning
  • Deep Active Learning – a mash-up of Deep Learning and Active Learning (see above), aimed at rendering Deep Learning more efficient (Note: There’s also a Deep Active Learning associated with Deep vs. Surface Learning.)
  • Feature Learning (Representation Learning) – in Machine Learning, any technique that requires a system to distill the defining features of a preselected set of representations so that it can use that information to analyze new data
  • Machine Teaching (2010s) – an aspect of the field of Machine Learning that is concerned with strategies to increase the relevance and utility of outputs. Most current efforts are focused on selecting and structuring training data, but attentions are also given to real-world applications and ongoing developing of algorithms and theories.
  • Probably Approximately Correct Learning (PAC Learning) (Leslie Valiant, 1980s) – an approach to Machine Learning in which the machine “learns” a concept by selecting appropriate functions and applying them to classify a set of samples. The aim is to achieve a high likelihood (the “probably”) of low errors (the “approximately correct”)
  • Reinforcement Learning – Drawing inspiration from Behaviorisms’ focus on rewarding desired behaviors and punishing undesired ones, in this approach algorithms run through many iterations of a multi-step task. Positive outcomes are incentivized and negative outcomes are penalized.
  • Self-Supervised Learning – This type of Machine Learning involves two steps: First, the system solves the task using pseudo-labels to sort unlabeled sample data. Second, the system performs the task with either Supervised Learning (see below) or Unsupervised Learning (see below).
  • Supervised Learning – This approach to Machine Learning is based on data mining. The machine is provided with a large set of examples, from which it is expected to determine a rule for inclusion and/or exclusion that can then be used to decide if new examples belong or don’t belong.
  • Unsupervised Learning – In this approach to Machine Learning , an untrained machine is provided with uncategorized examples and directed to group them according to similarities and differences.


At first blush, it might not seem appropriate to include Machine Learning in a survey of theories of learning associated with formal education. The decision to do so is justified in the fact that Machine Learning has become the source of important insights into and strategies to investigate human learning, enabled by a reversal of Cognitivism’s indefensible grounding metaphor (i.e., “brain as computer”) to an attitude that human cognition might be better understood through efforts to model and mimic the brain’s complexity and situatedness (i.e., “computer as brain”). That is, rather than positing a theory of learning based on an uninterrogated metaphor and, Machine Learning is affording insights into human cognition by attending critically to similarities and differences of brain-based and machine-based structures and dynamics.

Authors and/or Prominent Influences


Status as a Theory of Learning

Machine Learning is a round-about theory of learning – that is, it is a field of study that is affording profound insights into brain function and human learning through modeling and testing a broad array of machine-based strategies for perceiving, manipulating, sorting, and other tasks that are typically interpreted as cognitive.

Status as a Theory of Teaching

Machine Learning is not a theory of teaching, but findings are proving useful for informing basic matters of formatting and sequencing information to make it more accessible. (See Machine Teaching, above.)

Status as a Scientific Theory

Machine Learning is a robust domain of scientific inquiry.


  • Active Learning
  • Actor–Critic Model
  • Algorithmic Learning Theory (Algorithmic Inductive Inference)
  • Apprenticeship Learning (Learning from Demonstration)
  • Automated Learning
  • Deep Active Learning
  • Feature Learning (Representation Learning)
  • Machine Teaching
  • Probably Approximately Correct Learning (PAC Learning)
  • Reinforcement Learning
  • Self-Supervised Learning
  • Supervised Learning
  • Unsupervised Learning

Map Location

Please cite this article as:
Davis, B., & Francis, K. (2023). “Machine Learning” in Discourses on Learning in Education.

⇦ Back to Map
⇦ Back to List