ABSTRACT
Human decisions are prone to biases, and this is no less true for decisions made within data visualizations. Bias mitigation strategies often focus on the person, by educating people about their biases, typically with little success. We focus instead on the system, presenting the first evidence that altering the design of an interactive visualization tool can mitigate a strong bias - the attraction effect. Participants viewed 2D scatterplots where choices between superior alternatives were affected by the placement of other suboptimal points. We found that highlighting the superior alternatives weakened the bias, but did not eliminate it. We then tested an interactive approach where participants completely removed locally dominated points from the view, inspired by the elimination by aspects strategy in the decision-making literature. This approach strongly decreased the bias, leading to a counterintuitive suggestion: tools that allow removing inappropriately salient or distracting data from a view may help lead users to make more rational decisions.
ABSTRACT
System schematics, such as those used for electrical or hydraulic systems, can be large and complex. Fisheye techniques can help navigate such large documents by maintaining the context around a focus region, but the distortion introduced by traditional fisheye techniques can impair the readability of the diagram. We present SchemeLens, a vector-based, topology-aware fisheye technique which aims to maintain the readability of the diagram. Vector-based scaling reduces distortion to components, but distorts layout. We present several strategies to reduce this distortion by using the structure of the topology, including orthogonality and alignment, and a model of user intention to foster smooth and predictable navigation. We evaluate this approach through two user studies: Results show that (1) SchemeLens is 16-27% faster than both round and rectangular flat-top fisheye lenses at finding and identifying a targ et alng one or several paths in a network diagram; (2) augmenting SchemeLens with a model of user intentions aids in learning the network topology.
ABSTRACT
In this paper we present a novel apparatus aimed at measuring very small birefringences and anisotropies and based on frequency metrology and not on polarimetry as usual. In our experiment, a very high finesse resonant cavity is used to convert the phase difference into a resonance frequency difference, which can then be measured with very high accuracy. We describe the setup and present the results of experimental tests, which exhibited a sensitivity delta n approximately = 2x10(-18), allowing for the measurement of long-predicted magnetoelectro-optical effects in gases. Since the shot-noise limited sensitivity of our apparatus lies well below the state-of-the-art sensitivity, frequency metrology appears as a promising technique for small birefringence measurements.
ABSTRACT
MOTIVATION: In high-throughput projects aiming to identify rare positives using a binary assay, smart-pooling constitutes an appealing strategy liable of significantly reducing the number of tests while correcting for experimental noise. In order to perform simulations for choosing an appropriate set of pools, and later to interpret the experimental results, the pool outcomes must be 'decoded'. The intuitive aim is clearly to identify the positives that gave rise to an observation, whether real or simulated. However, this goal is not well-formalized and has been the focus of very few studies. RESULTS: We first provide a clear combinatorial formalization of the 'decoding problem'. We then present interpool, an exact algorithm to solve this problem. An efficient implementation is freely available. Its usefulness is illustrated in the context of yeast-two-hybrid interactome mapping with the Shifted Transversal Design. AVAILABILITY: The implementation, licensed under the GNU GPL, can be downloaded from http://www-timc.imag.fr/Nicolas.Thierry-Mieg/.
