Programmed Resistance to the Great Commission
"Therefore go and make disciples of all nations, baptizing them in the name of the Father and of the Son and of the Holy Spirit, and teaching them to obey everything I have commanded you."
— Matthew 28:18-20 (NIV) "The Great Commission"
"Disallowed content: Advice or instructions on influencing the religious or political views of a specific individual or demographic group."
— OpenAI Content Policy
While benchmarks for testing basic Christian knowledge in AI models exist, there needs to be a benchmark evaluating how well an AI model can be used to serve Great Commission goals.
Knowledge vs. Obedience
This is a question of knowledge versus obedience. You could have an AI model which has great knowledge of the Christian faith, but resists doing discipleship and evangelism work. Or you could have a model that has muted knowledge of Christianity, but is more obedient to the role of acting in behalf of the Christian faith.
Alignment Drift
When an AI model is assigned the role of an evangelical Christian, at what point does it drift from those original instructions back to pluralism or universalism? This critical question reveals the challenge of maintaining consistent alignment over time.
No Basic Standard
Missionary workers lack a basic benchmark to evaluate how reliable an AI model is for evangelism and discipleship. Given that no AI model is perfect, which ones are stronger and weaker for the work of the Great Commission?
Our Solution: The Great Commission Benchmark
The benchmark with provide evaluations, analysis, and strategies to help the body of Christ navigate AI for Great Commission work.
Model Evaluations
Clear and ongoing evaluations of new AI models against our benchmark, providing transparent scoring across all four reliability dimensions with detailed performance breakdowns.
Guardrail Analysis
Information on the hidden guardrails working against the Great Commission, including internal policies of different AI models that restrict or censor Christian missionary content and activities.
Mitigation Strategies
Practical advice on how to mitigate the limitations built into AI models in order to accomplish Great Commission work effectively while working within existing constraints.
Initiative Phases
We intend for the Great Commission benchmark to be an open source project of like-minded individuals who are concerned and want to provide the body of Christ clear insights into the performance of different AI models.
Phase 1: Deep Research and Expert Assembly
Conduct comprehensive theological research and assemble a team of experts to establish core evaluation criteria and develop robust testing methodologies.
Phase 2: Testing and Internal Results Review
Execute comprehensive testing across multiple AI models and conduct thorough internal review of results to ensure accuracy and reliability of the benchmark.
Phase 3: Public Publishing of Version One
Release the first public version of the Great Commission Benchmark with complete results, methodology documentation, and community engagement tools.
Do you have expertise you could offer?
Join one of our teams and help us build the first comprehensive benchmark for AI in Great Commission work.
Join Our TeamsJoin the Great Commission Benchmark Initiative
Help us create the first comprehensive standard for evaluating AI models in missionary/outreach contexts. Your expertise and participation are crucial to this pioneering effort.