In this study, I explore the learning opportunities that emerge from ill-structured activities in a university physics course and how collaboration enables performance under different teaching conditions. The study was conducted over two months on three sections of an introductory physics course in a University in Northern Chile. Each section utilized a different variety of teaching strategies and combinations of problems (well and ill-structured) for assessing content on the day of data collection, students were asked to work in groups and ill-structured activity in physics. I gathered audio of students' discussion in four groups for a total of 26 participants, and I collected the generated problems from the whole sample. From the audio, I explored the emergent processes students engaged while solving the problems. Student generated activities were coded to investigate combinations of concepts and problem characteristics, which were later combined into a measure of problem elaboration. Later, I explored students’ social networks to determine how different instructional strategies led to different social configurations and their differences in academic performance. For this, I collected data on students' performance on a physics test designed with well-structured problems and problem elaboration, and I asked students to respond to an on-line peer-nomination survey related to their social interactions engaged for information seeking to solve problems. I tested the effect of different network structures over academic performance on both types of activities by setting statistical linear models. Generating problems is an opportunity for students to propose ideas and make decisions regarding the content and the contextual details to introduce into their problems, as well as to engage in problem solving strategies. The combination of concepts and attributes for problem elaboration showed students' familiarity with particular portions of the content and characteristics, with differences across sections. Finally, students who actively sought out information from multiple peers were less likely to achieve good performance on well-structured problems, whereas for ill-structured problems, this effect depended on the features of the learning environment enacted in each section. These results suggest that teaching and instructional strategies have a key role in the way cooperation lead to good performance.