findAbar: how astronomers may perceive the bar in galaxies differently

TL;DR

Bars are common in galaxies, yet a universal, practical definition is lacking, risking biased interpretations of bar-related phenomena. The authors recruited 21 astronomers to classify 200 snapshots from two simulated barred galaxies, extracting $R_{ m bar}$, $b_{ m bar}$, axis ratio $b_{ m bar}/R_{ m bar}$, and bar orientation $\phi_{\rm bar}$, while tracking whether a bar is identified. They find substantial variation across observers in all bar parameters, with $\phi_{\rm bar}$ being the most robust and $R_{ m bar}$ and axis ratio showing notable discrepancies; gender and career stage emerge as significant biases. Automated bar-finders perform poorly relative to expert classifications, suggesting automation cannot yet replace human consensus and that an interim best practice—thoroughly documenting bar definitions with visuals and parameter tables—should be adopted. The work highlights how biases in bar classification can propagate into scientific results and advocates for community-wide standardization to improve repeatability in studies of barred galaxies.

Abstract

Bars are ubiquitous morphological features in the observed distribution of galaxies. There are similarly many methods for classifying these features and, without a strict theoretical definition or common standard practice, this is often left to circumstance. So, we were concerned whether astronomers even agree on the bar which they perceive in a given galaxy and whether this could impact perceived scientific results. As an elementary test, we twenty-one astronomers with varied experience in studying resolved galaxies and circumstances, have each assessed 200 galaxy images, spanning the early phase of bar evolution in two different barred galaxy simulations. We find variations exist within the classification of all the standard bar parameters assessed: bar length, axis-ratio, pitch-angle and even whether a bar is present at all. If this is indicative of the wider community, it has implications for interpreting morphological trends, such as bar-end effects. Furthermore, we find that it is surprisingly not expertise but gender, followed by career stage, which gives rise to the largest discrepancies in the reported bar parameters. Currently, automation does not seem to be a viable solution, with bar classifications from two automated bar-finding algorithms tested and failing to find bars in snapshots where most astronomers agree a bar must exist. Increasing dependence on machine learning or crowdsourcing with a training dataset can only serve to obfuscate any existing biases if these originate from the specific astronomer producing the training material. On the strength of this small sample, we encourage an interim best practice to reduce the impact of any possible classification bias and set goals for the community to resolve the issue in the future.

Figures (8)

• Figure 1: Upper: A standard face-on stellar density map for one of the barred galaxies included in findAbariles2022iles2024 with coloured ellipses tracing two different responses from the participating astronomers. Lower: An example of the image used for findAbar and the line selections for classifying the semi-major and semi-minor bar axes.
• Figure 2: The distribution of participants by career stage, gender and experience working with galaxies displaying bar structure.
• Figure 3: A subdivision of the distribution of participants indicating how the three participant attributes are distributed within the context of career stage.
• Figure 4: Representation of the full findAbar dataset in general parameters: $R_{\rm bar}$, $b_{\rm bar}$, Axis Ratio, $\phi_{\rm bar}$, and fraction of individuals who found a bar at each time-step. The bar parameter responses are presented by a boxplot for each snapshot. The central mark (yellow star) indicates the median response; the box spans from the 25th-75th percentile; whiskers (thin lines) extend to the most extreme data point within $Q_3+1.5(Q_3-Q_1)$ and $Q_1-1.5(Q_3-Q_1)$ and any values more extreme than these limits are classed as outliers (cross marks). The grey bars indicate the percentage of participants who identified a bar in a given snapshot, with the dashed line at 100% found a bar.
• Figure 5: The percentage difference between the median value for the responses with each participant attribute in each snapshot for each parameter (columns): $R_{\rm bar}$, $b_{\rm bar}$, Axis Ratio, $\phi_{\rm bar}$, and fraction of individuals who found a bar at each time-step. A value of zero indicates no difference in the median between the two attribute distributions. Positive value indicates that the attribute listed first is greater than the attribute listed second, while a negative value is the opposite.