Research

Topics

LLM

Experience Ratings

Reproducibility

Scale Compression

Intercoder Reliability

Open-Ended Responses

Survey Validation

Predictive Scoring

Text Analysis

Dimensionality in Satisfaction Ratings

LLMs decompose support conversations into satisfaction drivers, revealing lower customer satisfaction than surveys suggest.

Dimensionality in Satisfaction Ratings

LLMs decompose support conversations into satisfaction drivers, revealing lower customer satisfaction than surveys suggest.

Preview: The Dimensionality of Meaning

Our early validation work on text dimensionality provides key support for an emerging class of AI-native research methods.

Preview: The Dimensionality of Meaning

Our early validation work on text dimensionality provides key support for an emerging class of AI-native research methods.

What sentiment analysis can't see

Predicted satisfaction beats sentiment as a measurement of whether customers were helped, and what went wrong, across 70,000 support conversations.

What sentiment analysis can't see

Predicted satisfaction beats sentiment as a measurement of whether customers were helped, and what went wrong, across 70,000 support conversations.

The signal is the ceiling

Part 2: Reliability of AI annotation exceeds human benchmarks, however, the measurement limits of LLM-predicted ratings have an upper limit.

The signal is the ceiling

Part 2: Reliability of AI annotation exceeds human benchmarks, however, the measurement limits of LLM-predicted ratings have an upper limit.

LLM Predictive Scoring and Validation

Part 1: LLM-annotation recovers self-reported experience ratings from raw unstructured textual data across 10,000 MLB fan surveys.

LLM Predictive Scoring and Validation

Part 1: LLM-annotation recovers self-reported experience ratings from raw unstructured textual data across 10,000 MLB fan surveys.