Project summary

The VSPE Framework, short for Validation, Submission, Positivity, and Empowerment, is a four-step framework for safer, more emotionally intelligent AI programming.

This pilot delivers two things that make VSPE easy for vendors to adopt:

• Flattery-Reduction Benchmark – a proof-of-concept Colab notebook (20 Anthropic sycophancy prompts + 5 empowerment checks) that demonstrates ≥ 25 percentage-point drop in flattery responses.

• Licensing kit – plain-English contract template, tiered pricing, and a short slide deck so companies can license VSPE without heavy legal lift.

Related work: Varma & Beitman (2025) recently proposed a CBT-style “therapy loop” prompt to curb hallucinations. VSPE targets the complementary issue of flattery; our benchmark will include the therapy loop as a baseline for comparison.

What are this project's goals? How will you achieve them?

• Collaborate with a freelance ML engineer & licensing consultant; draft a 2-page benchmark/test plan (Aug 2025).

• Pitch assets – publish investor-grade slide deck, one-page summary, and polished website (Sep 2025).

• Benchmark demo – engineer ships the Colab; notebook shows ≥ 25 % flattery drop (Oct 2025).

• Legal & pricing – paralegal drafts non-exclusive license, pricing tiers, optional VSPE™ trademark filing (Nov 2025).

• Outreach & traction – contact 10 AI companies, run two user-feedback interviews, secure ≥ 2 expressions of interest by Jan 2026.

How will this funding be used?

Budget (total requested: $9,800)

• PI stipend – $3,000
  Five milestone payments of $600 for coordination, writing, and overall project management.

• Licensing consultant – $2,500
  Crafts plain-language contract and pricing tiers; includes Mia’s $1,200 advisor fee.

• Freelance ML engineer – $2,000
  About 15 hours at ≈ $130/hour to build and document the benchmark notebook.

• Legal & trademark search – $1,000
  Drafts the non-exclusive license and checks basic trademark availability for the VSPE Framework.

• Design & web polish – $500
  Updates the slide deck and landing page.

• User-feedback honoraria – $500
  Gift cards for two user-interview sessions / red-team reviews.

• Contingency buffer – $300
  Covers extra probes or—if unused—anything above $200 will be refunded.

*If only the minimum $6k is raised, we’ll still ship the benchmark, draft license, and slide deck. Trademark work and user interviews will move to a Phase 2 budget.

Who is on your team? What's your track record on similar projects?

Team & track record

• Astelle Kay (pen name for Kay Gwendolyn Astle) – Creator of the VSPE Framework. White paper found at my website: (https://www.vspeframework.com/p/vspe-a-psychologically-grounded-framework); provisional U.S. patent pending (#63/790,488 filed Apr 2025); advanced to Stage 2 of the 2025 MATS AI Alignment selection; graduate student with 4.0 GPA.

• Sergei Smirnov – Research Engineer & PhD student, University of Helsinki (AI alignment, mechanistic interpretability, evaluations). IEEE-published; 3+ years ML‐engineering experience; selected for the Finnish Alignment Engineering Bootcamp. Contracted for ~15 hrs to build and document the benchmark notebook.

• Licensing consultant / paralegal – to be recruited in Month 2; drafts license & pricing.

• Mia Abromitis – Associate Business Consultant, HealthEdge; leads claims-software implementations for healthcare payers and completed BlueDot Impact’s “Future of AI” course. Will provide two 30-minute compliance/ops reviews for VSPE.

What are the most likely causes and outcomes if this project fails?

• Benchmark effect < 25 % → iterate prompts, emphasize licensing narrative.

• Prompt over-submits → empowerment probes track any loss of helpfulness.

• API costs spike → switch to open-weights models (e.g., Mistral-7B).

• No EOI by Dec → publish results open-source and apply to LTFF for extended runway.

How much money have you raised in the last 12 months, and from where?

$0 – This is our second external funding request; the first one is pending.