Both Monitoring and Evaluation are often costly and time-consuming processes. According to Gail Campbell of the Zenex Foundation, M&E can consume between 10% to 30% of a project’s total budget. Moreover, in the context of evaluations that can span months or even years, Vindrola-Padros et. al (2021) emphasize that timeliness now plays a critical role in determining the extent to which M&E findings can genuinely inform decision-making processes. This underscores the importance of promptly delivering M&E insights to ensure their relevance and utility in decision-making. Considering the current prominence of AI in various domains, this blog investigates the potential of AI in enhancing the efficiency of monitoring and evaluation for complex health interventions.

What is AI, and does it have potential to revolutionize M&E processes?

AI, or Artificial Intelligence, can be defined as "the science of making machines do things that would require intelligence if done by people," as stated by NHS England. It encompasses the development of algorithms and models that enable machines to mimic or surpass human intelligence in specific tasks.

When considering the potential impact of AI in revolutionizing M&E processes, valuable insights can be drawn from the survey titled "Next Generation Professional" conducted by Devex and DAI. The 2018 survey, which involved over 2,500 development professionals, provided an important perspective on the influence of AI on M&E. Notably, approximately 25% of the participants expressed their belief that AI would be the most important technology area for the next generation, with the most significant impact on M&E, making it a field of global development that could undergo substantial transformation through the adoption of AI technologies.

The present application of AI in M&E 

The current application of AI in M&E encompasses two key areas: Natural Language Processing (NLP) and Computer Vision (CV), both of which hold significant promise for enhancing M&E practices.

NLP involves the use of AI algorithms to rapidly process, validate, and analyze qualitative data, such as textual information. It enables the interpretation of misspelled words from mobile-collected data, categorization of responses into qualitative categories (e.g., positive, negative, neutral), and rapid analysis of texts to derive valuable insights.

CV, on the other hand, focuses on AI models for facial recognition and object differentiation. Facial recognition can be applied effectively in attendance tracking for health interventions, enabling accurate estimation of the number of participants present at different timeframes. Additionally, sentiment analysis of photos can provide insights into human emotions, level of participation, engagement, and satisfaction. Furthermore, the object differentiation function of CV has been employed in supply tracking, where AI-based apps automatically verify delivered items in photos submitted by field staff, saving time and expenses during crisis events, as highlighted by Souktel Digital Solutions.

Limitations and ethics of AI 

In 2021, the World Health Organization issued a global report on AI with guiding principles for its design and use in healthcare and medicine. The report emphasized that systems trained primarily on data collected from individuals in high-income countries may not perform well for individuals in low- and middle-income settings. For example, NLP may struggle with complex sentences and texts in non-European major languages, while facial recognition in CV has shown biases towards white skin tones, posing challenges in identifying individuals with black skin. It also highlighted other technological, legal, security and ethical challenges which must be considered to ensure AI can fulfil its potential and promise. 

The future of AI in M&E

Looking ahead to the future, AI holds the substantial potential to revolutionize M&E of complex health interventions, offering cost and time savings. Current applications primarily focus on qualitative data processing, attendance monitoring, and supply tracking using object differentiation, each with its limitations. 

However, to bring about radical changes in traditional M&E practices, integration of AI and big data is essential. For example, AI-powered digital healthcare assistants like Dr. Elsa, when properly integrated with the relevant big data, have the potential to evaluate treatment efficacy across various health conditions. This offers a faster and more cost-effective alternative to traditional scientific experiments, enabling healthcare systems to assess the effectiveness of drugs and treatment procedures.