Research Track "Human-AI Decision Making"

Application of artificial intelligence in high stakes domains centers around human-AI decision making in which human stakeholders interact with AI. Examples in the financial sector include Know Your Customer (KYC) and Customer Due Dilligence (CDD) processes set in place to prevent banks from being used, intentionally or unintentionally, by criminal elements for money laundering activities.

Current advances in Natural Language Processing (NLP) open up new possibilities to strengthen KYC and CDD activities. Written documents from regulators describe the constraints and requirements banks and their customers must adhere to. The assessment of compliance is based on a series of data sources, which can be structured (stored in databases, queried by SQL) or unstructured, in which case it is, again, written plain text. The route explored in the context of AFR is to use NLP to distill the requirements from regulatory documents automatically. With that in place, customer-related data, can be used to assess compliance. Here, again, for the unstructured parts, NLP techniques need be applied to distill elements that can be used for formal compliance evaluation.

Selected Publications

• Sara Salimzadeh, Gaole He, Ujwal Gadiraju: A Missing Piece in the Puzzle: Considering the Role of Task Complexity in Human-AI Decision Making. UMAP 2023: 215-227 (preprint).

• Sara Salimzadeh, Ujwal Gadiraju, Claudia Hauff, Arie van Deursen: Exploring the Feasibility of Crowd-Powered Decomposition of Complex User Questions in Text-to-SQL Tasks. HT 2022: 154-165 (preprint)

• Sara Salimzadeh, David Maxwell, Claudia Hauff: The Impact of Entity Cards on Learning-Oriented Search Tasks. ICTIR 2021: 63-72 (preprint).