Assessment in social science should emphasise :

Modern social science pedagogy (as reflected in NCF 2005 and constructivist teaching-learning practice) treats assessment as a means of checking whether learners have meaningfully constructed and can use what they have learned — not merely a device for sorting learners, testing an isolated skill, or checking memory. Its primary purpose is to evaluate transfer of learning: whether a learner can apply concepts, values, and understanding drawn from social science to real, everyday situations.

Applying this to the given options: only evaluating learners' ability to extrapolate their learning to various life situations checks for this transfer and application of learning, which is exactly the competency-based, meaning-making purpose that social science assessment is meant to serve.

Contrasting with the other options:

• Identifying learners who are failing treats assessment as a sorting/labelling exercise rather than a tool for evaluating meaningful learning — assessment's role is broader than flagging failure.

• Testing writing ability narrows assessment down to one language skill rather than the learner's grasp and application of social science concepts.

• Measuring retention of information reduces assessment to rote recall — exactly the memorisation-based approach that contemporary social science pedagogy moves away from.

So assessment in social science should emphasise learners' ability to apply and extrapolate their learning to real-life situations, since this transfer of learning is what reveals genuine, usable understanding rather than rote or narrowly skill-based performance.