Mobile edge computing can reduce the processing burden on resource-constrained IoT devices, but shared servers introduce uncertainty and congestion. A cognitive approach uses partial offloading and risk-sensitive decision-making to balance local computation against shared edge resources.