Humans for AI
​Supporting AI systems to understand and reflect human nuance and reasoning in complex contexts
Human for AI systems
Bringing editorial experience and linguistics expertise to AI content evaluation and human reinforcement learning
I provide linguistic evaluation, values alignment assessment, and human feedback for AI systems.
This work draws on 30+ years of experience in communication science, linguistics, and behavioural neurobiology to help AI companies improve how their models understand and generate language.
​
I've spent decades helping people understand language. Now I'm extending that work to help machines do the same to contribute to the development of accurate, ethical AI systems.
​
​
Words for Change works with language at every level — from the fine detail of word structure to the broader strategy that shapes meaning in context.
Who it's for?
This work is for AI companies, research labs, and organisations building or developing AI systems who need:
-
Human evaluation of model outputs for quality, accuracy, and alignment.
-
Safety assessment and red-teaming from a linguistic and values perspective
-
Domain expertise in social, environmental, or educational contexts.
-
Evaluators who can spot subtle bias, framing issues, or cognitive patterns that undermine intended goals.
What's it for?
AI systems learn from patterns, but patterns don't capture everything that makes communication effective, ethical or trustworthy.
AI needs human evaluators who understand framing, cognitive patterns and nuance. Sound evaluation processes go beyond fact checking to assess linguistic and editorial accuracy because sometimes AI provides information that is technically correct but contextually wrong or expressed in an incorrect tone.
Types of evaluation work I provide:
Values alignment and safety evaluation
Assessing whether AI outputs reflect ethical principles, support human agency and avoid subtle harms. I can identify framing that undermines intended values, extractive metaphors, or subtle cognitive manipulation.
Red-teaming and bias detection
Stress-testing AI for problematic patterns. My background in framing, metaphor and values-based communication means I catch bias and misalignment that purely technical evaluation might miss.
Content review and quality assessment
Creating benchmark tasks, writing gold-standard solutions and evaluating AI-generated responses for clarity, accuracy. and comprehensiveness. This includes fact-checking, spotting inconsistencies and applying analytical judgment across diverse domains.
Domain-specific evaluation
Providing expert assessment in social, environmental, educational or policy contexts where nuance, values and communication effectiveness matter as much as technical accuracy.
Reinforcement learning from human feedback (RLHF)
Contributing human judgment to improve model outputs, particularly for complex domains where context, tone, framing and values alignment are critical.
Training data review
Assessing data for bias, problematic framing, or values issues before it shapes model behaviour.
What makes my evaluation skills set and approach different?
Most AI evaluators come from technical or data science backgrounds.
In contrast, and complementary to teams' tech and data skills, my interdisciplinary background offers:
​
Linguistic expertise: MA Linguistics, with deep knowledge of framing, metaphor, discourse patterns and how language shapes reasoning.
Cognitive science background: Training in behavioural neurobiology means I understand how people process information and make decisions.
Communication science experience: 30+ years working across research, education, science communication and values-based messaging
Editorial judgment: 15 years as an accredited copy editor, skilled in synthesis, clarity, and quality assessment across educational publishing (eg., senior science textbooks), academic editing for peer-review journals and scientific report writing and editing.
​
Systems thinking: PhD and Post Doc. in Ecology and interdisciplinary background means I evaluate for context, not just correctness. I spot things at multiple levels: semantic accuracy, pragmatic appropriateness, cognitive patterns, values alignment and strategic effectiveness.
Availability and approach
I work remotely and flexibly, with experience adapting to different evaluation protocols and project structures. My approach is rigorous, analytical and grounded in decades of professional practice across research, education, and communication.
​
I'm available for contract roles with AI companies, research labs, or organisations building human-centred, ethical AI systems.
Image: Marvin Meyer on UnSplash
Why Words for Change?
Applying the science of human communication to AI communication
Because understanding language is what I've done for 30 years. I've worked across ecology, linguistics, education, academia, science communication and values-based strategy.
I know how language works, how it fails, and what makes the difference between technically correct and genuinely good.
​
AI systems need evaluators who understand both the technical requirements and the human reality. That's the bridge I bring.
If you're building AI that needs to understand human nuance and reasoning, particularly in complex social, environmental, or educational contexts, let's talk.
