We propose a phenomenological model in which incoherent feedforward loops play a role in immune/tumor interactions, acting as "change detectors" and complementing mechanisms of self/non-self discrimination. There is a regulatory node X (level of activity of e.g. Treg population in tumor microenvironment), an effector node Y (immune activity, e.g. CD8+ effector T cells) repressed by X, and a node T (tumor load) which is (i) an excitatory input to both X and Y (IFFL) and (ii) repressed by Y. Without Y, T grows exponentially at a rate R. IFFL sensing leads to triggering of an immune response only if there is an acute change, but tolerance to constant, or even slowly varying, T. When autocatalytic feedback of Y is added (e,g, cytokine-mediated recruiting, autocrine stimulation), there results a bistable system, which can ``lock'' into a high state of activity in response to a transient rapid rate of change in its input. An an increase in the intrinsic growth rate R of the tumor, leaving all other parameters fixed, leads to elimination of T for small R, proliferation as R increases, but then (surprisingly!) elimination in the more aggressive case (when R is larger), followed of course by proliferation if R is even larger. An IFFL motif leads to predictions of logarithmic sensing (Weber phenomenon) and invariance (fold-change detection).

The very preliminary http://biorxiv.org/content/early/2015/12/30/035600 had more biological discussion, showing how our model is consistent with but greatly extends the tunable activation threshold model (Grossman Paul, PNAS1992), the discontinuity theory of immunity (Pradeu Jaeger Vivier, Nat Rev Immunology 2013), and the Tcell growth conjecture (Arias et al, Royal Soc 2015). A more detailed mathematical paper is in http://arxiv.org/abs/1602.00162