Cognitive Science AI Safety Research

Bridging cognitive science and AI to build safer, smarter systems.

PhD in Cognitive Neuroscience, University of Western Australia

I research how both biological and artificial systems process information—from EEG studies of working memory and ADHD to LLM safety evaluation frameworks.

View Work Collaborate
0
Years Research
0
Publications
0
Open Source

Profile

From cognitive development to AI safety research.

My research spans from understanding working memory and ADHD in children to developing frameworks for evaluating and securing large language models.

My journey began with EEG research on cognitive processes in children, investigating goal neglect and working memory. This foundation in neuroscience evolved into exploring how artificial systems process information and make decisions.

Today, I focus on AI safety and evaluation, developing tools like LLM Evals for systematic response grading and researching prompt injection vulnerabilities. I bridge academic rigor with practical applications.

EEG Analysis Machine Learning AI Safety Cognitive Science

  • Published research on goal neglect in children using EEG to understand cognitive control development.

  • Investigated working memory's role in ADHD symptoms through neurological and behavioral studies.

  • Developed LLM evaluation frameworks for systematic assessment of AI model responses and safety.

  • Created educational resources on cognitive science-based prompting at neuroprompting.xyz.

Cognitive Neuroscience

Brain Architecture

Optimizing sensory storytelling through live EEG sculpting. This lab builds audience-scale experiences where bioelectric rhythms manipulate light, particle systems, and spatial audio in real time.

Alpha Rhythm
10.5Hz
Beta Synchrony
22.3Hz
Gamma Flux
65.7Hz

Visitors don biometric headsets whilst the installation paints spectral pathways across a volumetric brain. Neural motifs respond to focus, relaxation, and surprise—offering a poetic biofeedback loop.

Enter Neuroscience Lab

Artificial Intelligence

Deep Learning Systems

I architect speculative and responsible machine intelligence—building shimmering networks that communicate trust through transparency, tactility, and play.

Layer Depth
12
Active Neurons
1,024
Model Accuracy
97.3%

Every simulation is paired with human-centred rituals—design critiques, ethics salons, and field tests—that keep emergent behaviour accountable.

Explore AI Systems

Signature Work

Building tools at the intersection of cognition and computation.

From research publications to open-source projects, my work applies scientific rigour to practical challenges in AI safety, evaluation, and human-computer interaction.

Prompt Optimiser

Interactive tool that enhances AI prompts using cognitive science principles. Applies neuroscience-based techniques to improve clarity, specificity, and effectiveness.

  • Cognitive Science
  • AI Optimisation
  • Interactive Tool
Try Tool
LLM Evals

Python library for systematic evaluation of large language model responses. Enables automated grading and quality assessment of AI outputs at scale.

  • Python
  • LLM Testing
  • Open Source
View Repo
Neuroprompting.xyz

Educational platform applying cognitive science principles to AI prompting. Teaches how human cognition insights can improve LLM interactions.

  • Cognitive Science
  • Prompt Engineering
  • Education
Visit Site
MiningRecords.com

Search and discovery platform for mining exploration data. Makes geological and resource information accessible through intelligent search.

  • Data Mining
  • Search
  • Web Platform
Explore

Research & Publications

Published research spanning cognitive development and AI safety.

My academic work bridges developmental psychology, neuroscience, and artificial intelligence, contributing to our understanding of both biological and artificial cognitive systems.

Journal of Attention Disorders

Working Memory and ADHD Symptoms

Investigated the relationship between working memory deficits and ADHD symptomatology in children, using neurological assessments and behavioral measures.

Learning and Instruction

Knowledge Transfer in Educational Contexts

Explored how cognitive load and instructional design impact learning outcomes and knowledge retention in educational settings.

AI Safety Research

Prompt Injection Vulnerabilities in LLMs

Analyzing security vulnerabilities in large language models, focusing on prompt injection attacks and developing defensive strategies.

Collaborator Voices

"An orchestra conductor for intelligence—equal parts scientist, artist, and ethicist."

"Gareth translated our dense neuroscience findings into a visceral installation that museum visitors still talk about. The empathy and rigor are unmatched."

Research Director · Cortex Lab

"The clarity Gareth brings to explainability transforms stakeholder anxiety into excitement. Our AI platform finally feels human."

Product Lead · Horizon Systems

"Workshopping speculative rituals with Gareth unlocked new policy perspectives for us. They bridge philosophy and prototyping seamlessly."

Policy Fellow · Future Commons

Interactive Experiments

Try these AI experiments and tools.

Interactive demonstrations exploring the boundaries between human and machine cognition.

Popular
Human or AI?

Can you tell who's real in a conversation? Test your ability to distinguish human responses from AI-generated text in this Turing Test game.

  • Pattern Recognition
  • AI Detection
  • Game
Play Now
Truth Arena

Watch AI agents debate topics with configurable honesty levels. Explore how deception and truth-telling manifest in generated arguments.

  • AI Debate
  • Deception
  • Experiment
Enter Arena
Prompt Optimiser

Enhance your prompts using cognitive science principles. Adjust clarity, detail, and creativity to craft more effective AI interactions.

  • Prompt Engineering
  • Cognitive Science
  • Tool
Try Tool

Let's Connect

Ready to collaborate on AI safety and cognitive science research?

Whether you're working on LLM evaluation, cognitive development research, or building safer AI systems, let's explore how we can work together.

Connect on LinkedIn Email Me