We assess the effects of our cellular manipulations using microscopy. However simply collecting images of cells undergoing various forms of manipulations is not enough on its own.
On a pragmatic level, a genome-wide screen can require 2,000,000 images to be captured; if you allow yourself a few seconds to look at each image (and allowed yourself to sleep occasionally) it would take almost 300 days to view all of the images and you are still unlikely to be able to come to any useful conclusion.
Thus we use image analysis techniques to convert raw images into numbers that have biologically relevant meaning. For any individual cell, typical measures could include, their shape, measures of intensity (amount of protein), number of fibres, spots and foci, size and shape of organelles. On top of that, as each well represents a population of cells, it is possible to measure population characteristics; how many cells are there, how close are they to each other, how heterogeneous the response of the population is to the treatment in that well etc.
This high content microscopy can thereby provide huge data description of the images without requiring the user to look at the images themselves. Moreover once captured, images can be easily and repeatedly reanalysed in silico providing a wealth of information for future researchers as knowledge and techniques for analysis evolve.