Start here ZIP · 6.57 MB
Complete companion
Chapter modules, all 23 notebooks, extension laboratories, supplied data, exercises and selected worked answers.
Download complete packageThe book & Python companion
Programming
Consciousness?
Models, Experiments, and Exercises with Python
Explore consciousness through computational models you can inspect, change and test.
Release 2026.11.03 Python 3.12 Runs locally
Forthcoming from World Scientific
Series on Machine Consciousness
THE EXPERIMENTAL APPROACH
- Chapters
- 18
- Notebooks
- 23
- Exercises
- 444
- Selected solutions
- 150
The question behind the code
From ideas
to experiments.
What does a machine do, and what would justify saying more?
How does a system select information, remember an instruction or assess its own uncertainty?
Programming Consciousness? approaches these questions through small Python models, controlled interventions and comparisons with rival explanations. Eighteen chapters connect perception, attention, memory and self-modeling with cognitive architectures, contemporary AI and responsible research.
Predict an outcome, inspect the mechanism, change an assumption and explain what the result supports. Most laboratories are controlled simulations; two analyse published human decisions and recorded software-agent runs.
Who this is for
Advanced undergraduate and graduate students in engineering and related fields who already know Python. Basic NumPy, probability, statistics and matrix operations are useful; no prior course in philosophy of mind or neuroscience is required.
A path through the subject
Explore the book
From a first inspectable model to an integrated architecture. Open a chapter to see its focus.
Showing all 18 chapters
Questions and experimental methods
01Python as an Experimental Language
Use explicit state, preserved traces and controlled changes to turn a Python program into an inspectable experiment.
02What It Means to Model Consciousness
Distinguish access, report and subjective experience, and clarify what a computational model can claim about each.
03From Theory to Experiment
Translate theoretical questions into predictions, controls and planned analyses while examining assumptions built into synthetic data.
Perception, attention and memory
04Perception and Decision Making Under Uncertainty
Use signal detection and psychometric models to separate sensory evidence, response criterion, decisions and visibility reports.
05Attention Selection and Competition
Compare selection mechanisms on shared inputs to examine competition, capacity and the difference between priority and task success.
06Feedforward and Recurrent Processing
Investigate recurrent dynamics and compare them with feedforward alternatives that receive the same observation history.
07Global Workspace and Broadcasting
Build a limited workspace and compare broadcasting, selective routing and local communication under explicit delivery budgets.
08Working Memory and Temporal Integration
Model admission, maintenance and retrieval in working memory, and test how interference and decision rules affect delayed performance.
Monitoring, causality and models
09Metacognition and Higher Order Representations
Separate decisions, confidence and control through synthetic experiments and held-out analyses of published human confidence reports.
10Information Integration and Causal Structure
Enumerate small Boolean networks to distinguish entropy, dependence, causal influence and a precisely defined partition-loss statistic.
11Prediction and World Models
Build a Bayesian filter, compare predictive baselines and examine how assumptions and information leakage change apparent forecasting success.
12Self Models and Attention Schema Theory
Test when a model of a selector’s errors changes action, using matched sensing budgets and a simpler rival controller.
Embodied agents, language and AI
13Embodiment Motivation and Affect
Model a situated agent whose resource needs guide action, and examine the functional roles of regulation and affect-like signals.
14Artificial Inner Speech
Compare verbal and nonverbal representations to test when an internal trace contributes to action and when it merely records it.
15Neural Networks Generative AI and Agentic Systems
Explore neural computation, attention and agent boundaries through small programs, recorded software-agent runs and an optional pretrained-model probe.
Comparison, responsibility and integration
16Comparing Competing Theories
Compare explicit models through predictions, selective interventions and independent synthetic observations, allowing ties and shared failures.
17Ethics and Responsible Research
Examine human responsibility, research safeguards and uncertainty about artificial welfare through an inspectable teaching protocol.
18A Theory Inspired Architecture and Bounded Self Modification
Integrate mechanisms in a seven-module architecture, test a viable rival and evaluate bounded policy revision under distribution shift.
The five themes above are a browsing guide. Chapter titles follow the book.
Teaching a course? Appendix E provides a twelve-week route with 36 core tasks.
Bring the laboratory with you
Download the companion materials
Release 2026.11.03
Choose the complete package, or the interface you prefer.
Scripts ZIP · 1.93 MB
Python laboratories
Runnable Python modules, required data and helpers, plus teaching materials. A direct route from the terminal.
Download Python packageInteractive records ZIP · 2.85 MB
Jupyter notebooks
All 23 notebooks with the code and data they need. Inspect a run, restart the kernel and make your own experiment.
Download notebook package
Exercise bank444 prompts with stable identifiersMarkdown
Selected solutions150 worked answers; not a complete answer keyMarkdown
Release READMESetup, package contents and source conditionsMarkdown
Third-party noticesData attribution, source links and source conditionsMarkdown
Download checksumsSHA-256 digests for checking file integrityText
Required calculations run locally with Python and NumPy on ordinary CPU hardware. No paid service, GPU or physical robot is required. The optional pretrained-model probe uses separate dependencies and a weights download; saved outputs are included. Source conditions and data attribution accompany the materials.
A first experiment
Run your first experiment
Use Python 3.12 and the companion release that matches your edition. Extract the archive, keep its folder structure intact and open a terminal in the extracted companion directory.
Begin with the release README. A virtual environment keeps the laboratory dependencies together.
Use python3 instead of python if that is how Python 3.12 is installed on your system.
01
Create and activate an environment
Create the environment
python -m venv .venvmacOS / Linux
Activate
source .venv/bin/activateWindows PowerShell
Activate
.venv\Scripts\Activate.ps1If your system prevents script activation, use .venv\Scripts\python.exe in place of python in the commands below.
02
Install the core dependencies
Python + NumPy
python -m pip install -r requirements.txt03
Run, inspect, then change one thing
First chapter / all chapter laboratories
python ch01.py
python run_all.pyCompare the trace with your prediction. Preserve the original result before changing a model, and record the release and environment used.
Prefer working in Jupyter?
Install the notebook dependencies in the same environment:
Optional notebook interface
python -m pip install -r requirements-notebooks.txtUse the Jupyter interface adopted by your course and select this Python environment. Open chapter notebooks from the companion directory and extension notebooks from their own folder. Restart the kernel and run cells in order.
Keep the record clear
Errata &
release notes.
No errata entries recorded yet for release 2026.11.03.
Check corrections against the edition and companion release you are using. The downloadable register provides a consistent place to track confirmed corrections.
Found something to correct?
Include the chapter and section or exercise number, the companion release, and a minimal example. For code, include the Python version, expected result and observed output.
Email Antonio Chella