papers/zoo-desci-platform/sections/05-ai-assisted-research.tex

\section{AI-Assisted Research}

\label{sec:ai}
%----------------------------------------------------------------------

\subsection{A-Chain Inference for Research}

\Zoo{} \AChain{} provides AI inference as a network service. For DeSci, this enables:

\subsubsection{Automated Literature Review}

A researcher submits a topic query to \AChain{}. The inference engine:

\begin{enumerate}
    \item Searches the Data Commons for relevant capsules (using embeddings stored on-chain).
    \item Retrieves and summarizes relevant findings from public Research Capsules.
    \item Generates a structured literature review with proper citations (linked to on-chain capsule IDs).
    \item Identifies gaps: topics with high query volume but few capsules.
\end{enumerate}

Cost: 500 $\KEEPER$ per literature review (inference credits).

\subsubsection{Hypothesis Generation}

Given a dataset and a research question, \AChain{} inference can:

\begin{enumerate}
    \item Analyze the dataset's statistical properties.
    \item Cross-reference with existing findings in the Data Commons.
    \item Propose testable hypotheses ranked by novelty and feasibility.
    \item Pre-register proposed hypotheses on \IChain{} (timestamped before experimentation).
\end{enumerate}

\subsubsection{Automated Statistical Analysis}

\AChain{} provides standardized statistical analysis pipelines:

\begin{itemize}
    \item Power analysis for sample size determination.
    \item Appropriate test selection based on data distribution and research design.
    \item Multiple comparison correction (Bonferroni, FDR).
    \item Effect size estimation with confidence intervals.
    \item Automated reporting in standardized format.
\end{itemize}

\subsection{Privacy-Preserving AI Analysis}

For sensitive data (medical records, endangered species locations), AI analysis runs on encrypted data via \CKKS{} on \AChain{} \cite{zoo-fhe}:

\begin{enumerate}
    \item The researcher encrypts the dataset under the collective \FHE{} key.
    \item \AChain{} runs the analysis model on encrypted inputs.
    \item Results are returned encrypted; only the researcher can decrypt.
    \item The analysis trace (encrypted) is anchored on \IChain{} for reproducibility.
\end{enumerate}

This enables AI-assisted analysis of sensitive data without the AI model (or its operators) ever observing the plaintext.

%----------------------------------------------------------------------