RCL to Use Artificial Intelligence for Expert Selection for the First Time
The Research Council of Lithuania (RCL), taking into account remarks from competent institutions regarding the expert evaluation process and seeking to ensure an even clearer, more transparent, and more consistent appointment of experts evaluating research proposals, is introducing an artificial intelligence (AI) solution into its expert selection process. The tool will be piloted for the first time in the Lithuania–Taiwan call administered by RCL, at the start of the expert selection stage.
The selection of experts is one of the most complex and sensitive stages of the evaluation process. Typically, this process takes around a month or longer, as it requires a thorough assessment of the proposal’s content, identification of scientists with appropriate expertise, verification of their experience and publications, and the prevention of conflicts of interest. In light of the complexity of the process and the associated risks, RCL has initiated a technological transformation of its expert selection procedures.
“Trust in the research funding system depends not only on final decisions, but also on the decision-making process itself. Therefore, we seek to ensure that expert selection is based on clear, verifiable, and consistently applied criteria. Piloting an artificial intelligence solution strengthens the robustness of our processes and helps reduce the risks of subjectivity and human error in large-scale evaluation stages. This is a purposeful step towards more data-driven and responsible decision-making,” says Dr Gintaras Valinčius, Chair of the Research Council of Lithuania.
The pilot AI tool is based on semantic analysis algorithms and advanced language models. The system converts a proposal abstract into a multidimensional mathematical vector and compares it with thousands of expert profiles. The assessment takes into account experts’ scientific experience, keywords, research areas, and actual publications indexed in international databases such as OpenAlex and Web of Science.
The system automatically calculates a relevance score for each expert and provides a reasoned shortlist of candidates. Additional mechanisms for reliability assurance, workload balancing, and conflict of interest (COI) prevention are applied – the system issues alerts if the match is insufficiently strong or if potential links are identified that may pose a risk of bias.
The technological solution enables the process to be carried out extremely quickly – for example, the tool will be able to process approximately 1,000 applications within 7–10 minutes. This would significantly shorten selection timelines, ensure greater substantiation of decisions, and further strengthen the transparency of the process.
The AI tool will function as an analytical assistant – a measure designed to support a more objective and systematic assessment of potential experts; however, final decisions will in all cases be made by a human.
Following the pilot phase, RCL will assess the tool’s effectiveness and accuracy. If successful, the solution will be applied to other calls with the aim of ensuring faster, more consistent, and more transparent expert selection.
RCL ensures the confidentiality of proposal content – when using the AI tool, proposal content is not transferred to external entities.