Observability in the world of Microsoft and AI
Observability
Azure AI Foundry supports Observability using some of the following evaluators:
Groundedness: measures how consistent the response is with respect to the retrieved context.
Relevance: measures how relevant the response is with respect to the query.
Fluency: measures natural language quality and readability.
Coherence: measures logical consistency and flow of responses.
Content safety: comprehensive assessment of various safety concerns.
So my word of the day is Groundedness, mostly because the “word” just jumped out at me as one of those new definitions for my brain, and in general vocabulary has other meanings for the growing generations.
In the Merriam-webster dictionary we find -> “Emotional and Psychological Context
It can also describe a person who is mentally and emotionally stable, realistic, and unpretentious. Someone with groundedness is often seen as calm, mindful, and not easily swayed by external chaos. [merriam-webster.com]”
and in Psychology Today we get “Spiritual and Mindfulness Context
Psychology Today defines groundedness as a sense of being fully embodied, whole, centered, and balanced in oneself and relationships. It’s associated with clarity, harmony, and a deep connection to the authentic self. [psychologytoday.com] “
So using the term as an evaluator trends towards a new sense of logic as we observe and evaluate AI results.
Field Notes 
Leave a comment