Toward a Neurosymbolic Understanding of Hidden Neuron Activations