Insight Beyond Numbers: The Impact of Qualitative Factors on Visual Data Analysis

High-resolution analysis of individual Drosophila melanogaster larvae uncovers individual variability in locomotion and its neurogenetic modulation

Neuronally orchestrated muscular movement and locomotion are defining faculties of multicellular animals. Due to its simple brain and genetic accessibility, the larva of the fruit fly Drosophila melanogaster allows one to study these processes at tractable levels of complexity. However, although the faculty of locomotion clearly pertains to the individual, most studies of locomotion in larvae use measurements aggregated across animals, or animals tested one by one, an extravagance for larger-scale analyses. This prevents grasping the inter- and intra-individual variability in locomotion and its neurogenetic determinants. Here, we present the IMBA (individual maggot behaviour analyser) for analysing the behaviour of individual larvae within groups, reliably resolving individual identity across collisions. We use the IMBA to systematically describe the inter- and intra-individual variability in locomotion of wild-type animals, and how the variability is reduced by associative learning. We then report a novel locomotion phenotype of an adhesion GPCR mutant. We further investigated the modulation of locomotion across repeated activations of dopamine neurons in individual animals, and the transient backward locomotion induced by brief optogenetic activation of the brain-descending 'mooncrawler' neurons. In summary, the IMBA is an easy-to-use toolbox allowing an unprecedentedly rich view of the behaviour and its variability of individual larvae, with utility in multiple biomedical research contexts.

Data Hunches: Incorporating Personal Knowledge into Visualizations

The trouble with data is that it frequently provides only an imperfect representation of a phenomenon of interest. Experts who are familiar with their datasets will often make implicit, mental corrections when analyzing a dataset, or will be cautious not to be overly confident about their findings if caveats are present. However, personal knowledge about the caveats of a dataset is typically not incorporated in a structured way, which is problematic if others who lack that knowledge interpret the data. In this work, we define such analysts' knowledge about datasets as data hunches. We differentiate data hunches from uncertainty and discuss types of hunches. We then explore ways of recording data hunches, and, based on a prototypical design, develop recommendations for designing visualizations that support data hunches. We conclude by discussing various challenges associated with data hunches, including the potential for harm and challenges for trust and privacy. We envision that data hunches will empower analysts to externalize their knowledge, facilitate collaboration and communication, and support the ability to learn from others' data hunches.

DashBot: Insight-Driven Dashboard Generation Based on Deep Reinforcement Learning

Analytical dashboards are popular in business intelligence to facilitate insight discovery with multiple charts. However, creating an effective dashboard is highly demanding, which requires users to have adequate data analysis background and be familiar with professional tools, such as Power BI. To create a dashboard, users have to configure charts by selecting data columns and exploring different chart combinations to optimize the communication of insights, which is trial-and-error. Recent research has started to use deep learning methods for dashboard generation to lower the burden of visualization creation. However, such efforts are greatly hindered by the lack of large-scale and high-quality datasets of dashboards. In this work, we propose using deep reinforcement learning to generate analytical dashboards that can use well-established visualization knowledge and the estimation capacity of reinforcement learning. Specifically, we use visualization knowledge to construct a training environment and rewards for agents to explore and imitate human exploration behavior with a well-designed agent network. The usefulness of the deep reinforcement learning model is demonstrated through ablation studies and user studies. In conclusion, our work opens up new opportunities to develop effective ML-based visualization recommenders without beforehand training datasets.