Cost-effectiveness and Digital Personalised Learning: Three interventions from Kenya

<ns3:p>Recently there has been a significant shift towards better understanding the relative cost-effectiveness of education interventions, particularly those implemented in LMICs. Work to group similar interventions has added value in identifying the types of programmes and forms of technology likely to deliver cost-effective improvements in learning outcomes. However, this framing relies on assumptions that programmes within the same category will have similar cost-effectiveness and be broadly similar in how they are implemented. Furthermore, it overlooks the key issue that individual programme features and their contextual implementation are consequential to cost-effectiveness. One category that has gained attention in recent years is Digital Personalised Learning (DPL) - a category that is difficult to define due to the variety of mechanisms through which learning can be personalised. As a result, it can be problematic to assume that the features and cost-effectiveness of DPL programmes can be compared and grouped, limiting conclusions about DPL as a category. This lack of consistency is emphasised by the absence of a universally agreed definition of DPL. Therefore, for understanding cost-effectiveness, it may be less useful to categorise all interventions with a personalisation element as comparable. In this paper, we use three examples of DPL interventions in Kenya to undertake an in-depth analysis of their features and implications for cost-effectiveness. By analysing key differences in implementation models, costs, and learning outcomes, this paper highlights significant variation in how DPL programmes achieve impact. It argues that superficial comparisons of cost-effectiveness may miss important information for decision-makers and educators comparing DPL options. As a result, DPL interventions should not be conceptualised as a single, discrete category when illustrating the cost-effectiveness of groups of interventions.</ns3:p>