Spiking neural networks (SNNs) are characterized by low power consumption, highly parallel computation and event-driven information processing [1]. Many of the concepts that promise to improve the efficiency of SNNs compared to conventional deep neural networks, such as spikes, sparsity, and static suppression, are inspired by our understanding of how the brain processes information. Hence, we can further advance the development of SNNs by improving our current knowledge of how biological neural networks learn through spike-based local plasticity mechanisms. In recent years, many spike-based learning rules have been proposed for training SNNs. However, assessing the biological validity of these rules remains challenging. Traditional approaches to validate plasticity paradigms, such as patch clamp or calcium imaging, are often limited by low throughput or the ability to record only from a subset of neurons within intricate networks. We present an in vitro plasticity model that allows us to study how spike-based plasticity mechanisms change synaptic weights to address these challenges. We leverage the relative simplicity and controllability of topologically constrained in vitro biological networks by engineering neural networks consisting of 2-3 human induced pluripotent stem cell-derived neurons within custom-designed polydimethylsiloxane microstructures [2]. Placing these structures on CMOS high-density microelectrode arrays enables us to collect data of pre- and postsynaptic neural activity. The system also permits to modulate activity of single neurons via localized stimulation. We obtain synaptic changes that last for multiple minutes when simultaneously stimulating pre- and post-synaptic neurons to induce Hebbian learning. The collected data will inform the development of a realistic in silico model to accurately reconstruct the in vitro network structure. The low complexity of our topologically constrained in vitro networks facilitates the reconstruction of functional behavior using biophysical models, which are limited by their high computational demands when modelling larger networks. These models will characterize synaptic changes in in vitro neural networks, facilitating the testing of various local learning rules and identifying plausible combinations. The biological validity of these learning rules can then be confirmed using the corresponding in vitro biological neural network, potentially inspiring new local learning rules.