fixed figure caption

paper/paper.md

@@ -59,7 +59,7 @@ Automated Research Assistant (`autora`) is a Python package for automating and i
 # Statement of Need
 The pace of empirical research is constrained by the rate at which scientists can alternate between the design and execution of experiments, on the one hand, and the derivation of scientific knowledge, on the other hand. However, attempts to increase this rate can compromise scientific rigor, leading to lower quality of formal modeling, insufficient documentation, and non-replicable findings. `autora` aims to surmount these limitations by formalizing the empirical research process and automating the generation, estimation, and empirical testing of scientific models. By providing a declarative language for empirical research, `autora` offers greater transparency and rigor in empirical research while accelerating scientific discovery. While existing scientific computing packages solve individual aspects of empirical research, there is no workflow mechanic for integrating them into a single pipeline, e.g., to enable closed-loop experiments. `autora` offers such a workflow mechanic, integrating Python packages for automating specific aspects of the empirical research process.
 
-![The `autora` framework. (A) `autora` workflow, as applied in a behavioral research study. `autora` implements components (colored boxes; see text) that can be integrated into a closed-loop discovery process. Workflows expressed in `autora` depend on modules for individual scientific tasks, such as running behavioral experiments, which are maintained by internal and external contributors. The core packages are actively maintained by the Autonomous Empirical Research Group. (B) `autora`’s components acting on the state object. The state object maintains relevant scientific data, such as experimental conditions X, observations Y, and models, and can be modified by `autora` components. Here, the cycle begins with an experimentalist adding experimental conditions $x_1$ to the state. The experiment runner then executes the experiment and collects corresponding observations $y_1$. The cycle concludes with the theorist computing a model that relates $x_1$ to $y_1$.\label{fig:overview}](figure.png)
+![The `autora` framework. (A) `autora` workflow, as applied in a behavioral research study. `autora` implements components (colored boxes; see text) that can be integrated into a closed-loop discovery process. Workflows expressed in `autora` depend on modules for individual scientific tasks, such as designing behavioral experiments, executing those experiments, and analyzing collected data. (B) `autora`’s components acting on the state object. The state object maintains relevant scientific data, such as experimental conditions X, observations Y, and models, and can be modified by `autora` components. Here, the cycle begins with an experimentalist adding experimental conditions $x_1$ to the state. The experiment runner then executes the experiment and collects corresponding observations $y_1$. The cycle concludes with the theorist computing a model that relates $x_1$ to $y_1$.\label{fig:overview}](figure.png)
 
 # Overview and Components
 The `autora` framework implements and interfaces with components automating different phases of the empirical research process (\autoref{fig:overview}A). These components include *experimentalists* for automating experimental design, *experiment runners* for automating data collection, and *theorists* for automating scientific model discovery. To illustrate each component, we consider an exemplary behavioral research study (cf. \autoref{fig:overview}) that examines the probability of human participants detecting a visual stimulus as a function of its intensity.