This paper investigates whether LLMs can be steered to generate explanations that are tailored to a specific group of people and proposes an approach that first identifies group-specific attributes in terms of explanatory style and knowledge of a specific target group and tailors the explanations significantly better to the target group while maintaining the best specificity-factuality balance.
Abstract
To effectively enable people to understand new topics, explanations should be tailored to their backgrounds and abilities. Prompting alone has been shown to be insufficient for creating such explanations, and other computational methods are missing so far. Therefore, this paper investigates whether LLMs can be steered to generate explanations that are tailored to a specific group of people. To this end, we propose an approach that first identifies group-specific attributes in terms of explanatory style and knowledge of a specific target group. Building on activation engineering, it then computes attribute-based steering vectors and adds them to the internal activations of an LLM during inference to enable a fine-grained steering. In our experiments, we assess the steering effectiveness in terms of specificity and factuality of the generated explanations. Additionally, we evaluate the explanations in a study with human experts from different target groups. Compared to prompting and state-of-the-art steering baselines, our approach tailors the explanations significantly better to the target group while maintaining the best specificity-factuality balance.
As interest in explainable artificial intelligence (XAI) continues to grow, a critical gap remains between the explanations generated by models and their interpretability by human users, particularly when cognitive biases and semantic relevance are not adequately addressed. This paper introduces CoRIfEE-Rel, a novel hu...
Parisa Mahya, Johannes Fürnkranz· Information· 0 citations
ConceptLens is a model-agnostic framework that transforms explanations into an interactive interface for probing pre-trained classifiers across domains and modalities, and enables users to evaluate local fidelity, perform what-if analysis, conduct counterfactual searches, and generate evidence-grounded explanations.
Hasan Md Tusfiqur Alam, Abdulrahman Mohamed Selim, László Kopácsi et al.· Proceedings of the 28th Inte...· 0 citations
It is proposed that observed human response variation of a survey question is a potential explanation for the mixed performance observed so far in persona prompting and new insights are provided on the effectiveness of different attribute selection methods for LLM-based survey prediction using persona prompting.
Leon Fröhling, Jens Rupprecht, Markus Strohmaier et al.· 3 citations
A systematic failure of selective control is identified and a disparity, the neutrality gap, demonstrates that successful persona following does not imply selective persona control, which additionally requires keeping non-target attributes stable.
Jia-Shen Ren, Wen-Lin Zhang, Bo-Han Zhang et al.· 1 citation
Comparing three model families and multiple math benchmarks, COMPASS outperforms the activation-steering baselines the authors compare against, improves GSM8K accuracy by 16 percentage points on average, and approaches CoT accuracy with 20-70\% fewer generated tokens.
Pratyay Dutta, Kowshik Thopalli, V. Narayanaswamy· 0 citations
It is demonstrated empirically that ProToMEx not only produces explanations of comparable fidelity to popular methods like SHAP and LIME but also drastically reduces the amortised computational cost of generating local explanations, making it highly suitable for real-time applications.
A. Georgara, Adarsh Valoor, Sarvapali D. Ramchurn· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 19, 2026
An MIT and Dana-Farber Institute-led team used large language models to examine nearly 4,000 studies and hundreds of code repositories, revealing gaps in the information needed to scrutinize and reuse published research.
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.