The Human/AI Language • Chapter 4

Ethics, Errors and Explainability

We are standing on the edge of a revolution. For the first time in history, machines can hold meaningful conversations with us, help us solve problems, and even inspire new ideas. But with great power comes great responsibility.

Every time you interact with an AI system, you’re shaping the future. The way you use these tools, the information you share, and the tasks you entrust to them don’t just affect your immediate outcomes—they influence how this technology evolves for society at large.

Estimated time.

Part 1

The Power of Your Choices

Every conversation with an AI language model helps shape tomorrow's technology. Your prompts aren't just questions and commands - they represent choices about AI's future role in society.

Consider how collective choices influence development:

  • How we respond to biased or inaccurate outputs signals what society demands from AI systems

  • Our verification practices help establish standards for AI deployment

  • Our choices about when to use AI help define its role in society

  • Market responses to AI capabilities influence future development priorities

  • Public discourse about AI ethics impacts regulatory and development frameworks

Key Principles:

  • Frame prompts that enhance rather than replace human thinking

  • Question responses that could perpetuate harm

  • Stay vigilant about accuracy and misinformation

  • Be mindful of impact on sensitive topics

  • Maintain human oversight

  • Consider broader societal impact when using AI

Today's AI systems might not learn directly from our individual conversations, but our collective choices - what we accept, what we reject, how and when we choose to use AI - these shape the future of the technology. Every interaction is a signal about the kind of AI we want to build.

Part 2

Privacy and Confidentiality

Language models process everything you share, using it to generate responses. While they don't store conversations permanently, understanding their data handling is crucial.

Before sharing information, consider:

  • Could this data affect someone's privacy, even indirectly?

  • Would I be comfortable with this information being used to generate responses?

  • Have I verified the system's privacy policies?

Never share:

  • Personal identifiers (NI number, passport details)

  • Financial credentials

  • Medical records

  • Confidential business data

  • Information about others

  • NDA-protected content

Best Practices:

  • Anonymise examples when seeking advice

  • Remove identifying details from scenarios

  • Consider creating synthetic data for testing

  • Default to sharing less, not more

Remember: Privacy isn't just about protecting your data - it's about respecting everyone's right to control their information in an AI-powered world.

Part 3

Staying Cautious and Critical

AI systems are incredibly powerful, but their responses can sometimes be misleading—either due to hallucinations (fabricated or incorrect information) or biases embedded in the data they were trained on. As a user, your role is to approach these outputs thoughtfully, questioning them and recognising when caution is needed.

Hallucinations

AI systems don’t say, "I don’t know." Instead, they generate the most plausible-sounding answer based on their training data, even when they lack sufficient information. This means they might confidently present incorrect details or fabricate facts entirely.

Best practices:

  • Verify outputs against reliable sources, especially for specific or technical content

  • Watch for overconfident or unusually detailed responses

  • Be sceptical of specific numbers, dates, or citations

  • Use follow-up questions to test consistency and reveal gaps

Bias

AI reflects the biases present in the data it was trained on, and these biases can subtly influence responses. For example, topics that dominate Western-centric media might receive more nuanced answers, while less-represented perspectives are over-simplified or entirely overlooked. Always remember that responses can reinforce stereotypes or cultural assumptions.

Best practices:

  • Frame questions to encourage balanced responses.

  • Specify desired perspectives or contexts in your prompts.

  • Question responses that oversimplify complex issues

  • Be alert to cultural assumptions and stereotypes

Remember, AI's confident tone doesn't guarantee accuracy. Maintain critical thinking and verify important information.

Part 4

Explainability, the Black Box of AI

Language models can't reliably explain their reasoning process. Unlike human experts who can walk you through their thought process, AI systems operate through complex pattern matching that isn't transparent.

While these models can seemingly explain their reasoning when asked, these explanations are effectively after-the-fact guesses rather than true insights into their decision-making process.

This creates two challenges:

  • Explanations may sound logical but not reflect the actual process

  • Plausible-sounding reasoning may be constructed for incorrect outputs

Best practices:

  • Request explanations for important responses

  • Verify reasoning against external sources

  • Break complex queries into traceable steps

  • Document your own decision-making process

  • Maintain human oversight for critical decisions

Remember, while AI can seemingly articulate reasoning, critical thinking and verification remain essential. Use AI to enhance, not replace, human judgment.

Reflection

Every interaction with AI is a choice that shapes our collective future.

As AI becomes more integrated into our lives, we face crucial decisions about how to engage with these powerful systems. The balance between leveraging AI's capabilities and maintaining human agency isn't just a personal choice - it's a societal imperative. How we protect privacy, verify accuracy, and maintain human judgment today will determine whether AI enhances or diminishes what makes us human tomorrow.

Your choices matter. They ripple outward, influencing not just how AI develops, but how humanity adapts to this transformative technology.

Next • Chapter 5

The future of LLMs

The pace of innovation in the field of LLMs is nothing short of extraordinary. Every new development pushes the boundaries of what these systems can do, offering glimpses into a future where humans and AI work together in ways we’re only beginning to imagine.

By exploring what lies ahead, we can prepare not just to adapt but to shape the future of AI communication.