The EPINET project will investigate conditions for the development of more integrated technology assessment (TA) methods. It will develop methods and criteria to be used for more socially robust and efficient practices on the interfaces between TA and the world of policy makers and innovators. At present, a large number of TA methodologies and practices exist. Many of these are based on varying – and sometimes conflicting, unclear - values, presuppositions, interests and commitments. This is problematic, insofar as differing conclusions and recommendations will follow from different methodologies and disciplines; hence the need for more integrated approaches. However, the irreducible difference of perspectives and plurality in the field of TA needs to be recognised and used as a resource. EPINET introduces the concept of epistemic networks as a way of conceptualising complex developments within emerging fields of sociotechnical innovation practices. It establishes a weak or “soft” framework within which the plurality of different TA practices can be explored in a concerted and holistic manner; EPINET uses this to study four cases: wearable sensors, cognition for technical systems, synthetic meat and smart grids. “Integrating TA”, it is claimed, is a task for empirical investigation in which implicit values of TA methodologies, disciplines and practices are spelled out and placed in relation to the practices they are meant to assess. This is the context of innovation conceptualised through the concept of emerging and future epistemic networks. EPINET develops a holistic framework for integrating assessments through gradual co-production of methodologies and concepts (centrally that of “responsible innovation”) together with innovators and policy makers. The challenges of “integrating assessments”, we claim, can only be gradually worked out within such a holistic view of complex intersecting networks and practices.