AI and Textmetrics: Getting the Most from Artificial Intelligence
Recently, Artificial Intelligence has captured the imagination of many in business because of its almost infinite possibilities and potential as a tool to benefit humanity. It may do everything from develop cures for previously incurable diseases, to develop impressive alternative energy sources. But for now, one of AI’s best, most approachable uses is as a tool for ideation. However, AI is also somewhat notorious for offering content that may include factual inaccuracies, implicit bias, and plain nonsense, sometimes referred to as “hallucinations.” As a tool, business must be able to rely on AI to produce consistent results, otherwise it’s of little use at all. What is needed is the ability to impose guardrails on the AI work product, giving it clear instructions on what to produce: the tone, the readability and reading level, context, style, credibility, checking for bias. What is needed is a mature, rules-based writing assistant, one that is not subject to the inconsistencies found in the content created by AI.
To improve the quality and consistency of AI’s output, we propose an integration model that is commonly used to implement user interfaces, data and controlling logic. This integration model separates the business logic and presentation layer from one another, giving reliable structure to the business logic results and predictability to how those results are presented to the user interface (UI) and, by extension, to its end users.
This integration isolates the application’s concerns into three distinct layers. The AI layer is responsible for the application’s logic, storing and retrieving data from back-end data stores. It will likely include mechanisms for validating data and carrying out other data-related tasks. The UI is necessary for the user to interact with the application. It displays the data and enables users to interact with it. The Textmetrics layer contains the business rules necessary to facilitate communications with clear, consistent content that can be relied upon to convey the organization’s preferred tone and style without bias.
AI is a Large Language Model (LLM), creating content with its extensive understanding of language patterns and knowledge. This LLM acts as the foundational element, processing and generating human-like text responses. On the user-facing side, the UI provides an intuitive interface for clients to engage with the language model. It ensures a user-friendly experience and facilitates meaningful interactions. Textmetrics interprets user inputs, leveraging Natural Language Processing (NLP) to understand context and intent, and guides the LLM to generate coherent and contextually relevant responses. Textmetrics acts as the orchestrator, harmonizing user interactions with the vast language capabilities of AI. And it is AI-agnostic, meaning that it can work with any AI, or quite well on its own without any AI input at all.
Because of extensive work with various iterations of AI, we have found that it operates with certain limitations. AI generates text based on patterns learned during training and doesn’t understand or generate text with specific emotional tones. While users may achieve a sense of tonality by shaping their inputs and contexts in a way that guides the model to respond in a desired manner, the model lacks an understanding of emotions or tonality; it is generating responses based on statistical patterns present in the data upon which it was trained. AI can exhibit a lack of genuine understanding, has a sensitivity to input phrasing, can be prone to bias (because of the implicit bias sometimes present in source data) and contextual dependence as well as exhibit difficulty handling complex tonal shifts.
Textmetrics gives structure to AI’s content creation by applying a detailed set of rules once, across all communications, so that the preferred tone and style of the generated content is consistent from one communication to the next, across all departments.
Send us your thoughts:
