Weak gravitational lensing observations are a key science driver for the NASA Wide Field Infrared Survey Telescope (WFIRST). To validate the performance of the WFIRST infrared detectors, we have performed a laboratory emulation of weak gravitational lensing measurements. Our experiments used a custom precision projector system to image a target mask composed of a grid of pinholes, emulating stellar point sources, onto a 1.7 micron cut-off Teledyne HgCdTe/H2RG detector. We used a 880nm LED illumination source and f/22 pupil stop to produce undersampled point spread functions similar to those expected from WFIRST. We also emulated the WFIRST image reconstruction strategy, using the IMage COMbination (IMCOM) algorithm to derive oversampled images from dithered, undersampled input images. We created shear maps for this data and computed shear correlation functions to mimic a real weak lensing analysis. After removing only 2nd order polynomial fits to the shear maps, we found that the correlation functions could be reduced to O(10^-6). This places a conservative upper limit on the detector-induced bias to the correlation function (under our test conditions). This bias is two orders of magnitude lower than the expected weak lensing signal. Restricted to scales relevant to dark energy analyses (sky separations > 0.5 arcmin), the bias is O(10^-7): comparable to the requirement for future weak lensing missions to avoid biasing cosmological parameter estimates. Our experiment will need to be upgraded and repeated under different configurations to fully characterize the shape measurement performance of WFIRST IR detectors. ; Comment: 57 pages (double-spaced preprint format), 17 color figures, accepted to PASP. Changes to version 2: fixed typos, figure quality