Conservation conflicts occur when people clash over conservation objectives. They are damaging for biodiversity, livelihoods, and human well-being globally, and are often managed via interventions intended to change people's behaviour. However, variation in intervention approaches across contexts remains underexplored. This thesis seeks to inform management by better understanding the roots of conflict, and the factors constraining the choice and efficacy of different interventions. Using an empirical literature review, I first identify five intervention types - 'technical', 'cognitive', 'economic', 'enforcement' and 'stakeholder' - and how they associate with conflict frames, behaviours and geography. I then largely corroborate these results in an experimental survey with conservation professionals, which also uncovers how decision-makers' characteristics, including disciplinary and demographic backgrounds, predict their intervention priorities. I then draw upon stakeholder interviews in Enduimet Wildlife Management Area in Tanzania, and grey literature to identify how multiple levels of conflict - covering human-elephant interactions, stakeholder interactions, and governance structures - constrain local management options. Next, using an experimental public goods game in Enduimet, I find that stakeholder perceptions of intervener trustworthiness predict levels of cooperation with conflict interventions. Lastly, by analysing conflict over trophy hunting via the social media platform Twitter, I reveal how this issue is polarised along similar political and value-based dimensions as other environmental conflicts. Beyond advocating for behaviourallyinformed interventions, these findings have three key management implications: that the backgrounds of decision-makers mediate their priorities, that the backgrounds of interveners mediate responses to interventions, and that the socio-political and governance background of conflicts will likely mediate the outcome of interventions.