There is a need for a more comprehensive sociotechnical systems view on ML. Such a view looks at the development and use of an ML system in practice as being a sociotechnical ML system: "a system consisting of technical artefacts, human agents and institutions, in which a machine-based subsystem influences its real or virtual environment by automating, supporting or augmenting decision-making". This research takes on this view to design a sociotechnical guide for ML practice, centring the specication of sociotechnical ML systems. Taking on the guidelines contributes to a safe and effective development and use of ML systems.

The authors present the application of a retarding field between the electron objective lens and sample in an integrated fluorescence and electron microscope. The retarding field enhances signal collection and signal strength in the electron microscope. This is beneficial for samples prepared for integrated fluorescence and electron microscopy as the amount of staining material added to enhance electron microscopy signal is typically lower compared to conventional samples in order to preserve fluorescence. We demonstrate signal enhancement through the applied retarding field for both 80-nm post-embedding immunolabeled sections and 100-nm in-resin preserved fluorescence sections. Moreover, we show that tuning the electron landing energy particularly improves imaging conditions for ultra-thin (50 nm) sections, where optimization of both retarding field and interaction volume contribute to the signal improvement. Finally, we show that our integrated retarding field setup allows landing energies down to a few electron volts with 0.3 eV dispersion, which opens new prospects for assessing electron beam induced damage by in situ quantification of the observed bleaching of the fluorescence following irradiation. @en

Large-scale electron microscopy (EM) allows analysis of both tissues and macromolecules in a semi-automated manner, but acquisition rate forms a bottleneck. We reasoned that a negative bias potential may be used to enhance signal collection, allowing shorter dwell times and thus increasing imaging speed. Negative bias potential has previously been used to tune penetration depth in block-face imaging. However, optimization of negative bias potential for application in thin section imaging will be needed prior to routine use and application in large-scale EM. Here, we present negative bias potential optimized through a combination of simulations and empirical measurements. We find that the use of a negative bias potential generally results in improvement of image quality and signal-to-noise ratio (SNR). The extent of these improvements depends on the presence and strength of a magnetic immersion field. Maintaining other imaging conditions and aiming for the same image quality and SNR, the use of a negative stage bias can allow for a 20-fold decrease in dwell time, thus reducing the time for a week long acquisition to less than 8 h. We further show that negative bias potential can be applied in an integrated correlative light electron microscopy (CLEM) application, allowing fast acquisition of a high precision overlaid LM-EM dataset. Application of negative stage bias potential will thus help to solve the current bottleneck of image acquisition of large fields of view at high resolution in large-scale microscopy.@en