Reconfigurable Intelligent Surfaces (RIS) stand as a pivotal technology in the realm of next-generation wireless communications, offering programmable radio propagation environments. Comprising multiple reflecting elements, RIS enable optimization of their phase and amplitude to attain desired reflection characteristics. However, the passive reflecting nature of RIS poses substantial challenges for channel estimation, making coherent symbol detection a formidable task. In modern communication systems, meeting the escalating throughput demands entails equipping Base Stations (BS) with multiple antennas for simultaneous communication with numerous users, enhancing spectral efficiency in Multi-User Multiple-Input Multiple-Output (MU-MIMO) systems. Symbol detection emerges as a key challenge in such systems, necessitating the retrieval of multiple symbols transmitted over the uplink channel. In this paper, we investigate the uplink of RIS-aided MU-MIMO systems and propose a low-complexity algorithm for symbol detection. We derive a simple data detection model using the virtual angular domain representation and precoding for data symbols, followed by the introduction of an iterative symbol detection (ISD) algorithm based on the symbol-by-symbol Maximum A Posteriori (MAP) criteria. Our algorithm reduces computational complexity by incorporating the finite nature of the data alphabet and addressing the sparsity of the sensing matrix through the virtual angular domain representation.