People aren’t blank slates. While the old idea, famously critiqued by Steven Pinker in his book The Blank Slate, suggested we start with empty minds, ready to be filled, that’s not quite right. Pinker argues that humans are born with inherent traits, instincts, and predispositions that significantly shape our learning and behavior. So, while our experiences influence us, they don’t write on an empty slate but rather interact with our inborn capacities.
AI systems, on the other hand, truly start from zero. They have to be taught everything, from basic knowledge to complex interactions. This includes how to understand and communicate with people effectively, which isn’t an inborn skill but one that is meticulously coded and trained.
It’s not just loading them with hard facts like science and Shakespeare. We also need to imbue them with soft skills. Most importantly, these include how to communicate and interact with people — skills that are often more nuanced and harder to teach than factual knowledge.
Many companies instill these communication skills by teaching their AI the three H’s: honest, helpful, and harmless. This means programming AI to always strive to provide truthful information (honest), to ensure their interactions are beneficial to users (helpful), and to avoid causing any harm or offense (harmless). These guidelines help make AI interactions more pleasant and safe.
I thought, “these are good principles for being a better communicator.” The CEO of Anthropic has a talk labelled Helpful, Honest, Harmless AI. But there is a more specific version of these guidelines used by people on the autism spectrum, who might struggle with standard social cues and norms. To navigate social interactions more smoothly, they may employ specific filters or questions to guide their responses and interactions. One of these filters looks like this:
- Is it truthful? This involves assessing whether the information being communicated is accurate and honest. It helps avoid misunderstandings and provides a solid foundation for trust.
- Is it helpful? This checks if the response or statement adds value to the conversation or assists the person they are interacting with. It promotes constructive and supportive interactions.
- Is it kind? This ensures that the communication is considerate and mindful of others’ feelings, aiming to prevent hurtful words or actions.
By examining how we program AI to interact in human-like ways, we can learn a lot about our own communication practices. It can highlight what makes communication effective and show us how to apply these techniques to improve our own interactions.
You must be logged in to post a comment.