This page provides access to several models for AI-supported coding of negotiation transcripts. Each is described below, including the definitions of the codes, and examples of each code. Also included are details of how the AI model was developed and validated, as well as instructions for how to format and submit your transcripts for coding.

These models were trained in most cases by learning from transcripts that have been human coded for specific prior projects. Each project used a different coding scheme, and even when codes were the same, they may have been operationalized slightly differently. The example sentences allow you to see how each project used each code, so you can decide if that model fits your research needs.

Costs: Feel free to try a few transcripts. If you plan to code large numbers of transcripts, we ask you to cover the fees we pay Anthropic to run the models. See this document to calculate an estimate of the cost. Contact us if you have any questions. In addition, note that NTR provides grants to cover the cost of these fees. To find out more, and to submit for a grant, see this page on NTR’s website.

For a summary of different strategies we used to develop these models, and why we chose the current strategy, see the following paper:

Link To PDF: PDF of paper

Meeting Journal Requirements. For your coding to be used in journal publications, they usually require information about the model, such as data security and privacy, the prompt used, the use of third-party LLMs, and validation data. This document should provide all the information needed for publication. Note that one of the premier journals for publishing negotiation research, Organizational Behavior and Human Decision Processes, has published an announcement that AI is acceptable to use for coding here.