Source-linked AI summary
A Collection of Definitions of Intelligence
Shane Legg, Marcus Hutter
TL;DR
The paper examines the lack of a standard definition of intelligence and surveys more than 70 informal definitions collected from diverse sources. It organizes these definitions across collective, psychologist, and AI-researcher formulations, finding recurring emphasis on learning, adaptation, problem solving, and goal achievement.
Problem
Intelligence has no standard definition despite a long history of research and debate.
Method
The authors compile and organize more than 70 informal definitions from collective sources, psychologists, and AI researchers.
Results
The collection shows recurring themes including learning, adaptation, problem solving, mental abilities, and goal achievement across definitions.
Takeaways & Limitations
Many proposed definitions are strongly related and share common features, supporting comparison across formulations rather than reliance on a single standard definition.
Abstract
from arXiv · showhide
This paper is a survey of a large number of informal definitions of ``intelligence'' that the authors have collected over the years. Naturally, compiling a complete list would be impossible as many definitions of intelligence are buried deep inside articles and books. Nevertheless, the 70-odd definitions presented here are, to the authors' knowledge, the largest and most well referenced collection there is.
1 Introduction
The paper addresses the absence of a standard definition of intelligence by surveying recurring similarities across proposed definitions. It presents more than 70 informal definitions as a large, well-referenced collection.
- Despite extensive research and debate, intelligence still lacks a standard definition.
- The authors argue that many definitions share strong similarities and may express the same idea in different words.
- The paper presents more than 70 informal definitions collected over the authors’ years of research.
- The collection is described as the largest and most well-referenced one known to the authors.
2 Collective definitions
Collective definitions portray intelligence as a broad capacity involving learning, reasoning, understanding, knowledge application, problem solving, and adaptation. The section attributes definitions to groups, organizations, dictionaries, and encyclopedias where no single psychologist is the source.
- Collective definitions are limited to unattributed or collectively agreed formulations, with dictionary sources traced toward original sources where possible.
- Across sources, intelligence commonly includes learning, understanding, reasoning, and applying knowledge or skills.
- The collective formulations also include broad mental abilities such as planning, problem solving, abstract thought, language, memory, and judgment.
- Several definitions characterize intelligence as adapting effectively to environments or dealing with new situations and problems.
3 Psychologist definitions
Psychologist definitions depict intelligence as a composite or general mental capacity expressed through learning, judgment, abstraction, problem solving, adaptation, and behavior. Some definitions connect it explicitly to practical life, culture, or psychometric measurement.
- Psychologists variously define intelligence as a composite of functions, a general capability, or a capacity underlying thought and problem solving.
- Adaptation to environments and novel situations is a recurring theme, including learning from experience and reorganizing behavior effectively.
- Other definitions emphasize acquiring, storing, retrieving, combining, and applying information in new contexts.
- Some formulations tie intelligence to culturally valued achievements or performance on intelligence tests.
- Intelligence is also described through abstract thinking, judgment, reasoning, imagination, and purposeful planning.
4 AI researcher definitions
AI researchers commonly define intelligence through adaptive, goal-directed behavior under environmental uncertainty and limited knowledge or resources. Their formulations emphasize achieving goals across environments, learning, and appropriate information processing.
- Several formulations make uncertainty, incomplete knowledge, finite computation, or limited resources part of intelligent action.
- AI definitions characterize intelligence as a system’s ability to achieve goals through appropriate, adaptive behavior in varied environments.
- Some definitions focus on solving hard problems, finding adequate solutions in large search spaces, or performing well across broad task ranges.
- Information-processing accounts define intelligence as adapting properly in complex environments, with criteria of properness acquired through processing.
- Intelligent systems are described as improving over time, learning from experience, or acquiring knowledge and skills autonomously.
5 Is a single definition possible?
The authors find that many intelligence definitions share common features, despite no objectively correct single definition. They therefore adopt an informal definition centered on achieving goals across varied environments.
- Intelligence is treated as a property of an agent interacting with its environment or environments.
- It is related to an agent’s ability to succeed or profit with respect to a goal or objective.
- Intelligence depends on the agent’s ability to adapt to different objectives and environments.
- The authors combine these attributes into the definition that intelligence measures an agent’s ability to achieve goals in a wide range of environments.
- Learning, adapting, and understanding are implicit because these capacities enable success across a wide range of environments.