Project summary

Elo Clínico is an AI alignment framework that engineers a "Dialectical Partner" for physicians, using phenomenological fine-tuning and SEM-PLS validation to prevent "phenomenological blindness" and empathy erosion in medical diagnostics.

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

Goal: To cure "clinical psychopathy" in Medical AI. Current models predict diagnoses but ignore pathos (subjective suffering).

Execution:

1. Phenomenological Fine-tuning: We are training LLMs on high-fidelity clinical datasets to detect emotional cues and prompt doctors with critical inquiries (e.g., "The patient’s syntax suggests isolation; have you assessed social support?").

2. Statistical Validation: We use Structural Equation Modeling (SEM-PLS with 5,000 bootstrap resamples) to mathematically prove that these humanistic variables act as protective factors against psychiatric hospitalization.

3. The 'Anomaly' as Lab: The project reverse-engineers my own cognitive reconstruction after a 2024 cerebellar stroke to design these empathy metrics.
   

How will this funding be used?

• $30,000 (Technical Layer): High-memory compute (H100s) for fine-tuning the "Dialectical Partner" model and acquiring specific phenomenological datasets.

• $10,000 (Operational Layer): Logistics for the Phase 1 Pilot at UFMG, deploying the interface in real-world primary care settings under Ethics Approval.

• $10,000 (Stipend/Artifact): Buy-out of my clinical time to focus 1.0 FTE on research and the final production of A QUIMERA (the project's philosophical source code).
  

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

João Medeiros da Fonseca (Lead Researcher): I am an "Anomaly": a synthesis of 10+ years in elite cultural journalism (Folha de S.Paulo), medical training at UFMG, and advanced statistical mastery.

• Track Record: I have successfully modeled the impact of Primary Care attributes on hospitalizations using public health data (SEM-PLS).

• Creative Proof: I secured a publication deal for A QUIMERA with Laranja Original, endorsed by an original book cover design from avant-garde legend Guto Lacaz.
  

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

Likely Cause: "Friction in the Clinical Loop." Overworked doctors in the public system (SUS) might reject a "Dialectical Partner" that asks them to pause and reflect, preferring faster, reductive automation. Outcome if Failed: The "product" deployment fails, but the research succeeds: we will have created the first "Phenomenological Safety Benchmark" for Medical AI in the Global South, which will remain a valuable open-source asset for future alignment researchers.

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

  $0. To date, this project has been fully bootstrapped through personal resources and the academic infrastructure of the Federal University of Minas Gerais (UFMG).