The empirical reconstruction of ecological interaction networks from data has remained a long-standing challenge. Recently, the debate on hypergraphs has emerged as a means to explain these complex systems. In this work, we aim to take a step further by integrating higher-order interactions (HOIs) into ecological networks derived from experimental time-series data. We proceed first by fitting a pairwise model using the generalized Lotka-Volterra (gLV) framework, and later expanding it to produce a model incorporating interactions between three species, which we refer to as triadic gLV. We compare our results against two different null models: one focusing on network link redistribution (null model A) and another on the randomization of time series prior to inference (null model B). Using various metrics, we quantify the differences between the networks inferred from the original data and the ensembles of networks generated under the two null models. Our findings indicate that pairwise networks are statistically less dense than their null model counterparts and characterized by a higher proportion of negative interactions. Conversely, networks incorporating HOIs are more populated than the null expectations and exhibit a high percentage of triadic interactions, despite showing a persistent preference for pairwise links. We find evidence of a structure governed by simplicial closure, where triadic interactions involving only two of the three possible underlying pairwise links appear to be suppressed, while those with all three are significantly overrepresented. We observe a correlation between degree and hyperdegree distributions, which the temporal randomization null model manages to capture. This master’s thesis provides deeper insight into the organizational strategies of ecological networks under a hypergraph approach and highlights the need for further experimental study of these interaction signs.