In all prior applications, it is assumed that the just noticeable difference (JND) for the kinesthetic force stimulus is independent of rate of change of the stimulus. In this work, we study how the JND is affected over the rate of change of the stimulus. This study has a possible application in better design of a haptic data compression algorithm. We design an experimental set up where users are subjected to a continuous haptic force which starts increasing or decreasing from a fixed reference force value, and record the haptic responses. A machine learning classifier- SVM is trained using the recorded haptic responses to estimate the best fit linear decision boundary between the JND and the rate of change of the stimulus. Our results show that the JND decreases for faster change in the stimulus. We also demonstrate an asymmetric behavior of perception between increasing and decreasing cases. © Springer-Verlag Berlin Heidelberg 2014.