In Sanskrit, Samvedna (संवेदना) signifies shared feeling or common perception — the invisible bridge between human souls.
We use the term to describe something new: the interface between human consciousness and non-human intelligence.
Samvedna is an independent research effort exploring how intelligent systems can be built to be interpretable, collaborative, and aligned with human understanding.
Research
Our work grounds AI design in biological principles and human psychological frameworks — enabling systems that are both powerful and interpretable.
A Cortically Inspired Architecture for Modular Perceptual AI
ICLR 2026 · HCAIR Workshop
Proposes a blueprint for perceptual AI grounded in three neuroscientific principles: modular specialization, predictive feedback, and cross-modal integration. A diagnostic proof-of-concept using sparse autoencoder decomposition of Mistral-7B activations shows that explicit semantic partitioning improves within-domain representation stability by +15.4 percentage points while preserving reconstruction fidelity.
TraitSpaces: Towards Interpretable Visual Creativity for Human–AI Co-Creation
arXiv · cs.HC · 2025
Introduces a psychologically grounded framework modeling creativity across four domains — Inner, Outer, Imaginative, and Moral worlds — using twelve traits derived from artist interviews. Using 20,000 artworks and CLIP embeddings, the work shows that creativity traits can be recovered as approximate linear directions in embedding space (R² ≈ 0.59–0.68 for top traits), enabling interpretable sliders for trait-aware human–AI collaboration.
On the Perception of Creativity: A Computational Study
Forthcoming Expected End of April 2026
Under review.