Editorial Statement

Editorial Statement

Alexandra M. Newman
Graduate Program in Operations Research with Engineering
Department of Mechanical Engineering
Colorado School of Mines
Golden, Colorado 80401
[email protected]

• IJAA encourages submissions of manuscripts with new and impactful applications in established areas such as transportation, logistics, and supply chain management, as well as energy systems, healthcare, electronic commerce, and the sharing economy. Similarly, we welcome mature methods in optimization, queueing theory, and decision analysis, as well as emerging techniques in data science, especially when they advance data-driven decision making. We solicit papers that enhance the journal’s visibility, including literature reviews and tutorials.
• In addition to the special issues featuring the Edelman Award and Wagner Prize finalists, we look forward to your ideas for special issues. We are always looking for practice summaries and relevant book reviews. Please see the "Calls for Papers" page for more information and to find any current special issues accepting submissions.
• We seek submissions from authors in countries that have traditionally been underrepresented in the journal and corresponding submissions addressing real-world challenges that operations research and analytics have been able to address in those countries.

With the proliferation of artificial intelligence, both in technical work and in writing, the INFORMS Journal on Applied Analytics recognizes the potential of large language models to enhance the application of analytics and broaden access to advanced analytics tools. We therefore welcome and encourage submissions that include the usage of LLMs in analytics’ application. However, LLMs also introduce new and distinct challenges. Unlike conventional models, which operate within more well-defined mathematical frameworks and for which established validation mechanisms exist, LLMs (which are, at their core, probabilistic next-token generators) create context-dependent responses that may be opaque, difficult to validate, and prone to plausible but incorrect information. These characteristics necessitate additional safeguards to ensure responsible use, especially when LLM outputs are consumed directly by decision makers without expert oversight.

To address these concerns and uphold the journal’s commitment to rigor and practical relevance, authors submitting manuscripts that involve LLMs in decision-making pipelines must include the following:

1. Validation Mechanisms
   Clearly describe how the validation of the LLM’s outputs is carried out. This includes:
• Techniques to detect and mitigate hallucinations or erroneous outputs.
• Methods and capabilities for verifying the correctness and relevance of model-generated recommendations.
• Any human-in-the-loop processes or automated checks implemented.
2. Transparency and Interpretability
   Detail if and how any decision maker that accesses or utilizes the LLM is informed about the model’s limitations, uncertainty, and potential failure modes. If explanations are generated by the LLM, discuss if and how their accuracy and reliability are assessed.
3. Decision Maker Responsibility
   Acknowledge that the final responsibility for decisions lies with the decision maker. Describe how the system supports informed decision-making, including safeguards against over-reliance on the model.
4. Ethical and Practical Considerations
   Discuss any ethical implications of deploying LLMs in the described context. Highlight any training, documentation, or support provided to end-users.