UCLA

Complex systems in science can be represented using networks and can be better understood by studying network-based principles that bridge different scientific disciplines. This dissertation showcases three studies that exemplify the use of network data to answer distinct scientific questions. Chapter 1 introduces network concepts and briefly discusses the scientific background and rationale for each of the three studies that make up the dissertation. Chapter 2 evaluates the differences between cooperation, defection, and punishment decisions in two series of online network games using decision times and assessed if experimental time pressure could shift the distribution of decision types away from punishment and toward cooperation; results show that punishment was slower than either cooperation or defection and that experimental time pressure did not reduce the frequency of punishment comparing games with and without time pressure. Chapter 3 uses an agent-based framework to simulate a network intervention to increase the lung cancer screening rate in the United States and tests two modifications to the framework with the potential to further increase the screening rate (allowing agents to adopt the intervention prior to observation and expanding the network size); results show that the intervention increases the screening rate among eligible individuals to 51% after 10 years of simulation (compared to the current screening rate among eligible individuals of 5.8%) and that increasing the network size boosts the screening rate above the rate achieved by the intervention alone. Chapter 4 describes the use of data from a citation network of scientific articles from epidemiology and public health to evaluate potential biases in ChatGPT’s ability to rate the validity of a scientific citation; results show that citation-reference pairs that included cited articles from English-speaking or high-income countries were rated higher than pairs with cited articles from non-English-speaking or low-income countries. Chapter 5 summarizes the results of the three studies and describes implications of those results for future research. Overall, this dissertation demonstrates the use of network science principles and network data to explore solutions to three challenges from behavioral science, intervention planning, and artificial intelligence.