What is PeopleAnalyst?

PeopleAnalyst is the front door for people-analytics research: 205+ works indexed and profiled, 40+ citation-grade findings extracted, and peer-reviewed behavioral science translated from academic to actionable — the missing manual for the people analytics you always meant to do.

What is people analytics?

People analytics is not a dashboard. It is behavioral science and statistical inference applied to workforce decisions — a discipline with its own methodology, spanning measurement, organizational design, talent, leadership, and analytics craft.

Why does AI in HR need measurement science?

AI is being deployed in high-stakes people decisions — hiring, performance, attrition — without the measurement science to evaluate whether it works or whom it harms. Construct validity, effect sizes, and criterion validity are the vocabulary for asking an AI vendor the right questions.

How is the research made accessible?

The evidence is indexed and searchable: 205+ works, 40+ citation-grade insight cards, and 8 research arcs, so the right finding reaches the right decision at the right time.

What separates good people measurement from assertion?

Good measurement has a method: construct validity, reliability, and effect-size interpretation are not optional — they are what separates evidence from assertion.

library / lib072479899414bafb

Artificial Intelligence - A Very Short Introduction

Very Short Introductions

In a sentence

A concise, expert tour of what artificial intelligence is, how its major approaches work, what it can and cannot do, and what its philosophical and social implications are.

Margaret Boden, a founding figure in cognitive science, demystifies artificial intelligence by framing it as the project of making computers do the sorts of things minds can do. She maps the field's five major approaches—symbolic (GOFAI), neural networks, evolutionary computing, cellular automata, and dynamical systems—and explains the unifying idea of the 'virtual machine': the information-processing system that matters, not the hardware. Across chapters on general intelligence, language, creativity, emotion, neural networks, robotics, artificial life, the philosophy of mind, and the Singularity, Boden shows that AI has produced spectacular narrow successes while artificial general intelligence remains a distant Holy Grail. She argues that AI's greatest lesson is how astonishingly rich human minds actually are, and she replaces hype with careful, sceptical judgement about both AI's promise and its real present-day dangers. Anyone wanting a rigorous yet accessible orientation to the conceptual foundations and stakes of AI will find it here.

The four lenses

Science
Statistics
Systems
Strategy

Tags

The model

A framework model inferred from Boden's argument: design choices in AI (method selection, virtual-machine sophistication, knowledge representation, tractability strategies, hybridization), conditioned by computational resources and biological inspiration, drive intermediate processing capacities (relevance/common-sense handling, learning, generalization) that determine outcomes such as narrow-task performance, progress toward AGI, and societal impact.

AI Method Selectiondesign lever

The choice among the major families of AI techniques—symbolic/GOFAI, neural networks, evolutionary programming, cellular automata, and dynamical systems—and their combination, tailored to the task at hand.

Virtual Machine Sophisticationpsychological state

The informational power and structural richness of the virtual machine (the information-processing system) implemented, independent of the underlying hardware, including its layered and hybrid organization.

Computational Resourcescontextual condition

Available hardware power, memory, data availability, and growth following Moore's Law, which enable but do not by themselves guarantee advanced virtual machines.

Tractability Strategiesdesign lever

Techniques for making problems solvable in practice—heuristics, planning, mathematical simplification, and efficient search ordering—that direct or shrink the search space.

Knowledge Representation Qualitydesign lever

The adequacy of how a problem and its domain knowledge are encoded—rules, frames, semantic nets, logic, word-vectors, or distributed networks—for enabling the system to reason and act effectively.

Biological Inspirationcontextual condition

The degree to which AI designs draw on real brains, organisms, evolution, and self-organization (e.g., situated robotics, GasNets, reaction-diffusion, neuromorphic computing).

Learning Capacitybehavioral pattern

The system's ability to improve through supervised, unsupervised, reinforcement, or deep learning, including discovering multilevel representations from data.

Relevance and Common-Sense Handlingpsychological state

The capacity to judge what is relevant, apply everyday knowledge, and avoid the frame problem—the central bottleneck distinguishing narrow systems from genuinely general intelligence.

Architectural Hybridizationdesign lever

The integration of symbolic and connectionist (and other) processing within a single system to combine complementary strengths, including whole-mind architectures.

Narrow-Task Performanceoutcome metric

Measurable success of AI on specialized tasks such as game playing, image recognition, translation, planning, and expert advice, often matching or exceeding humans.

Progress Toward General Intelligenceoutcome metric

The degree to which systems approach integrated, broad, common-sense, human-level intelligence spanning reasoning, perception, language, creativity, and emotion.

Societal Impact and Riskoutcome metric

The real-world consequences of AI deployment, including productivity gains alongside unemployment, privacy erosion, autonomous weapons, and deceptive emotional companions.

How they connect

method selection → influences narrow task performance
method selection → influences virtual machine sophistication
computational resources → moderates virtual machine sophistication
virtual machine sophistication → predicts progress toward agi
tractability strategies → influences narrow task performance
knowledge representation quality → influences narrow task performance
knowledge representation quality → influences relevance and common sense
learning capacity → predicts narrow task performance
biological inspiration → influences learning capacity
hybridization → influences virtual machine sophistication
relevance and common sense → predicts progress toward agi
narrow task performance → influences societal impact
computational resources → correlates progress toward agi

The process

This book provides a conceptual playbook for understanding and creating artificial intelligence, presenting AI not as a single technique but as a diverse collection of information-processing paradigms. The overall approach is to select the appropriate method from a toolbox of major strategies—symbolic, connectionist (neural networks), and evolutionary—based on the specific problem to be solved. The playbook begins with foundational, logic-based methods like classical planning, which structure problems in terms of goals, actions, and logical steps, a hallmark of "Good Old-Fashioned AI." It then moves to data-driven, biologically-inspired approaches, detailing the core methods of machine learning. These include supervised learning for when labeled data is available, unsupervised learning for discovering patterns in unlabeled data, and reinforcement learning for training systems through trial-and-error with rewards. This highlights a shift from explicitly programmed logic to systems that learn and adapt on their own. Finally, the playbook includes evolutionary programming, a method for generating novel and optimized solutions by simulating processes of variation and selection. This approach is suited for complex design and problem-solving where the solution space is vast and not fully understood. Together, these processes form a comprehensive overview of the fundamental workflows used to build systems that perceive, reason, learn, and create, illustrating the multi-faceted nature of the field of Artificial Intelligence.

Solving a Problem with a Classical Planner

To devise a sequence of actions to achieve a specific goal in a structured environment, breaking down a complex problem into manageable sub-problems.

When to use: When a task requires a sequence of steps to be performed to reach a desired outcome, such as in logistics, robotics, or solving puzzles like the 'missionaries-and-cannibals' problem.

Step 1Represent the problem by defining the final goal, a list of possible actions (operators), and the necessary prerequisites for each action.
Entry: The problem must be formalizable into goals, states, and actions.
Exit: A complete and accurate representation of the problem space is created.
In: Problem definition, Goal state, Set of possible actions · Out: Formal problem representation
Step 2Construct a plan at a suitable level of abstraction, specifying a sequence of actions as a hierarchy of goals and sub-goals.
Entry: The problem space is formally represented.
Exit: A high-level plan or goal hierarchy is established.
- Choose which goal or sub-goal to work on next.
- Select an appropriate action to achieve the current goal.
In: Formal problem representation · Out: Goal hierarchy, Abstract plan
Step 3For a given goal, select an action and verify that its prerequisites are met.
Entry: A goal has been selected from the plan hierarchy.
Exit: An applicable action is identified and its prerequisites are satisfied or set as new sub-goals.
- Does the action's prerequisite need to be a new sub-goal?
In: Current goal, List of actions · Out: Selected action, New sub-goals (if necessary)
Step 4Assemble the sequence of actions into a final plan.
Entry: All goals and sub-goals in the hierarchy have been resolved into executable actions.
Exit: A complete, ordered sequence of actions is produced.
In: Resolved goal hierarchy · Out: Final plan

Training a System with Supervised Machine Learning

To train a system to classify inputs or predict outcomes by learning from a large set of labeled examples.

When to use: When you need to build systems for tasks like face recognition, speech recognition, or medical diagnosis based on past data.

Step 1Prepare a training dataset containing a range of inputs, each paired with a desired, correct outcome (a label).
Entry: A clear classification or prediction problem is defined.
Exit: A large, well-structured, labeled dataset is available.
In: Raw data, Problem definition · Out: Labeled training dataset
Step 2Present inputs from the training data to the learning system.
Entry: The learning system and labeled dataset are ready.
Exit: The system has processed an input and produced an output.
In: A single labeled data point · Out: System's predicted output/classification
Step 3Compare the system's output to the correct, desired outcome from the training data to identify any error.
Entry: The system has produced an output for a given input.
Exit: An error signal is calculated.
In: System's predicted output, Correct output from labeled data · Out: Error signal
Step 4Provide feedback to the system to amend its internal hypothesis and reduce the error.
Entry: An error signal has been calculated.
Exit: The system's internal parameters (e.g., weights) are updated.
In: Error signal · Out: Adjusted internal model
Step 5Repeat the process with many examples from the training set until the system's performance on new, unseen data reaches a satisfactory level.
Entry: The system is capable of iterative learning.
Exit: The system achieves a target accuracy level on a separate validation dataset.
In: Entire training dataset · Out: A trained model

Discovering Patterns with Unsupervised Machine Learning

To enable a system to find hidden patterns, structures, or clusters within a dataset without any pre-existing labels or desired outcomes.

When to use: When you have a large amount of data and want to discover knowledge or groupings within it that you are not already aware of.

Step 1Provide the system with a large, unlabeled dataset.
Entry: A dataset is available for analysis.
Exit: The system has access to the complete dataset.
In: Unlabeled dataset
Step 2The system processes the data, driven by the principle that co-occurring features engender expectations that they will co-occur in the future.
Entry: The system has access to the data.
Exit: The system has analyzed the relationships between features in the data.
In: Unlabeled dataset · Out: Internal representation of data structure
Step 3The system finds and outputs the inherent patterns or clusters that exist in the data.
Entry: The system has analyzed the data's structure.
Exit: A set of discovered patterns, clusters, or associations is produced.
In: Internal representation of data structure · Out: Identified patterns or data clusters

Optimizing Behavior with Reinforcement Learning

To train a system (an agent) to make optimal decisions by interacting with an environment and receiving feedback in the form of rewards or punishments.

When to use: When designing a system that must learn to operate in a complex, dynamic environment through experience, such as learning to play a video game or controlling a robot's movements.

Step 1The agent performs an action or a sequence of actions within its environment.
Entry: An environment and a set of possible actions for the agent are defined.
Exit: The agent has executed an action and changed the state of the environment.
- Which action should be taken in the current state?
In: Current state of the environment · Out: An action
Step 2The environment provides feedback to the agent in the form of a reward or punishment.
Entry: The agent has performed an action.
Exit: A reward signal is sent to the agent.
In: Outcome of the action · Out: Reward signal (e.g., numerical score)
Step 3The agent uses a credit assignment procedure to determine which decisions were most responsible for the outcome.
Entry: A reward signal has been received.
Exit: Responsibility for the reward is attributed to specific prior actions.
In: Reward signal, History of recent actions · Out: Credit-assigned actions
Step 4The agent updates its internal strategy or policy to make actions that lead to higher rewards more likely in the future.
Entry: Credit has been assigned to prior actions.
Exit: The agent's policy for selecting actions is updated.
In: Credit-assigned actions · Out: Updated policy
Step 5Repeat the cycle of action, feedback, and policy update until the agent's performance converges on an optimal strategy.
Entry: The agent is capable of continuous learning.
Exit: The agent's performance reaches a plateau or a desired level of competence.
Out: An optimized agent policy

Developing a System with Evolutionary Programming

To create novel or highly optimized solutions to complex problems by simulating the process of biological evolution, including random variation and non-random selection.

When to use: When seeking to design a system, such as a robot controller or an electrical circuit, where the optimal design is unknown and can be improved through iterative trial and error.

Step 1Create an initial population of candidate solutions (programs or designs).
Entry: A problem has been defined that is amenable to an evolutionary approach.
Exit: An initial, diverse population of candidate solutions exists.
In: Problem definition · Out: Initial population of programs/designs
Step 2Define a 'fitness function' that provides a clear, quantitative criterion of success.
Entry: An initial population exists.
Exit: A computable fitness function is defined.
In: Performance requirements · Out: Fitness function
Step 3Apply genetic algorithms (GAs) to introduce random changes into the population, creating a new generation.
Entry: A population and fitness function are defined.
Exit: A new generation of candidate solutions with random variations is created.
In: Current population · Out: New, varied population
Step 4Evaluate each solution in the new generation using the fitness function.
Entry: A new generation has been created.
Exit: All members of the population have a fitness score.
In: New population, Fitness function · Out: Scored population
Step 5Select the most successful solutions to 'breed' the next generation.
Entry: The population has been scored.
Exit: A new parent population is selected for the next cycle.
- Which selection strategy to use (e.g., elitism, tournament selection)?
In: Scored population · Out: Parent population
Step 6Repeat the cycle of variation, evaluation, and selection over many generations.
Entry: The process is iterative.
Exit: A solution is found that meets the desired performance criteria, or the process is halted after a set number of generations.
Out: Optimized or novel solution

The story

The reader A curious, intelligent general reader who wants a clear, trustworthy understanding of what AI really is and what it means for minds, machines, and society.

External problem

AI is everywhere and rapidly advancing, but its concepts, capabilities, and limits are obscured by jargon and hype.

Internal problem

The reader feels uncertain, alternately dazzled and alarmed, unsure what to believe about AI's promise and its threats.

Philosophical problem

Public discourse that confuses narrow success with general intelligence and treats speculation as fact does a disservice to honest understanding.

The plan

Grasp the core idea that AI is about virtual machines doing what minds do.
Learn the five major AI approaches and how each processes information.
See why general intelligence, language, creativity, and emotion are so hard.
Examine the philosophical questions about real intelligence, consciousness, and morality.
Weigh the Singularity sceptically and focus on AI's genuine present-day risks.

Success

The reader can distinguish narrow from general AI and hype from substance.
The reader understands the major methods and their respective strengths and limits.
The reader can reason clearly about AI's philosophical and social stakes.
The reader holds a balanced, sceptical-yet-informed view of AI's future.

At stake

Remaining captive to misleading hype or unfounded dread about AI.
Mistaking impressive narrow performance for genuine understanding or general intelligence.
Overlooking the real, immediate risks—unemployment, privacy, autonomous weapons, deceptive companions.

Questions this book answers

What is artificial intelligence, and what are its scientific and technological aims?
What are the major types of AI and how do they process information?
Why is general intelligence (AGI) so much harder than narrow, specialist AI?
Can AI genuinely model language, creativity, and emotion?
Could an AI system have real intelligence, understanding, consciousness, freedom, or moral standing?

Glossary

AI Method Selection: The deliberate choice among major AI paradigms and their combination to suit a given task.
Virtual Machine Sophistication: The informational power and structural richness of the implemented information-processing system, abstracted from hardware.
Computational Resources: The hardware power, memory, and data availability supporting AI systems.
Tractability Strategies: Techniques that render problems solvable in practice by directing or reducing the search space.
Knowledge Representation Quality: The adequacy of how domain knowledge and problems are encoded for effective reasoning and action.
Biological Inspiration: The extent to which AI designs model real biological structures and processes.
Learning Capacity: The system's ability to improve performance through experience and data.
Relevance and Common-Sense Handling: The capacity to determine relevance, apply everyday knowledge, and avoid the frame problem.

Related in the library

Tools these methods power

Related in the literature

The measurement literature behind this signal — sourced, so you can defend it.

“Title : Artificial Intelligence: A Very Short Introduction (Very Short Introductions) Author: Boden, Margaret A. ASIN : B07DPP4J1B ISBN : 9780191080074 Artificial Intelligence : A Very Short Introduction VERY SHORT INTRODUCTIONS are for anyone wanting a stimulating and…”
— Artificial Intelligence a Very Short Introductionmatch 74%
“Kurzweil, R. (2005), The Singularity is Near: When Humans Transcend Biology (London: Penguin).Kurzweil, R. (2008), The Age of Spiritual Machines: When Computers Exceed Human Intelligence (London: Penguin).Bostrom, N. (2005), ‘A History of Transhumanist Thought’, Journal of…”
— Artificial Intelligence a Very Short Introductionmatch 67%
“All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed…”
— Artificial Intelligence a Very Short Introductionmatch 63%

Resources: Artificial Intelligence a Very Short Introduction