What is Dialogue Engineering?
Working with AI can be a science rather than an art. As our understanding of AI grows, so do our techniques of handling it. We believe that Dialogue Engineering is the next evolution of AI wrangling.
Background
When ChatGPT was first released, developers were quick to utilise it for production systems. In order to effectively use the technology, engineers have created Prompt Engineering. The concept was that an effective prompt could generate a desirable result. Prompt Engineering still remains the baseline of good AI usage, and it is how we interact with AI through chatbot interfaces.
As AI became more proficient and capable, engineers could begin integrating tool-calling and agentic workflows. However, an AI system has a finite memory, so developers were quick to realise that memory management was critical. Thus, Context Engineering was born. Whenever you interact with an AI, there are various processes running in the background to ensure that the AI can work as efficiently as possible.
Dialogue Engineering
However, we as users do not have much control over context management. This means that we are at the whims of deciphering AI outputs. We also have to restart the context if the AI outputs begin to veer from our desired outputs. This is obviously not ideal.
Dialogue Engineering is a step back and trying to understand why we interact with AI systems in the first place.
A dialogue is a conversation between you and yourself, you and the information, and you and the AI. As the user, you should have control over how that conversation is managed.
Anthrodontics has created various functions for managing this conversation.
- Re-writing your AI prompts.
- Deleting irrelevant parts of the conversation.
- Adding cells in arbitary positions to increase the context
- Re-arranging the order in which the conversation is structured.
- Editing the LLM output.
- Deleting undesired LLM ouputs