Agentic SHIFT: A Multi-Agent AI Framework for Visual and Context-Aware Cultural Heritage Exploration

Historic urban destinations contain culturally significant places whose digital visibility, accessibility, and suitability for visitation vary substantially. This paper presents Agentic SHIFT, the Smart Heritage Itinerary Framework for Transmission, a context-aware multi-agent framework for auditable cultural heritage itinerary recommendation. Nine specialised agents coordinate visitor profiling, context modelling, heritage retrieval, operational eligibility, heritage and under-exposure scoring, expected suitability, pedestrian routing, evidence-bounded narration, and visual verification. The Marrakech benchmark contains 196 cultural POIs across ten heritage categories. Twenty calibration configurations were used to select the utility weights, and forty held-out paired configurations were used for the final comparison. On the held-out set, Agentic SHIFT achieved mean heritage coverage of 0.757, digital under-exposure of 0.767, cultural diversity of 0.522, expected suitability of 0.631, and a mean walking distance of 1.973 km. Relative to popularity ranking, it increased heritage coverage by 0.047 and digital under-exposure by 0.448 while reducing walking distance by 0.375 km. Removing the operational gate reduced operational validity from 1.000 to 0.696. Two stored OSMnx routes confirmed pedestrian-network execution within a four-hour budget. A documentary audit of thirteen narrative claims found eleven directly supported claims and two claims retained with explicit qualification, while a detailed audit of four visual records confirmed identity, creator, licence, and publication-suitable resolution for all four audited images.

Authors

Institutions

Publication Details

Journal
Journal on Computing and Cultural Heritage
Published
2026-11-16
DOI
https://doi.org/10.1145/3849870
Primary Topic
Recommender Systems and Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Agentic SHIFT: A Multi-Agent AI Framework for Visual and Context-Aware Cultural Heritage Exploration

John K. C. Kingston, Sara Qassimi, Omar Bencharef, Abderrahmane Samali et al.
Journal on Computing and Cultural Heritage
Recommender Systems and Techniques
article

Agentic SHIFT: A Multi-Agent AI Framework for Visual and Context-Aware Cultural Heritage Exploration

John K. C. Kingston, Sara Qassimi, Omar Bencharef, Abderrahmane Samali, Abdessamad Imider
article en

Abstract

Historic urban destinations contain culturally significant places whose digital visibility, accessibility, and suitability for visitation vary substantially. This paper presents Agentic SHIFT, the Smart Heritage Itinerary Framework for Transmission, a context-aware multi-agent framework for auditable cultural heritage itinerary recommendation. Nine specialised agents coordinate visitor profiling, context modelling, heritage retrieval, operational eligibility, heritage and under-exposure scoring, expected suitability, pedestrian routing, evidence-bounded narration, and visual verification. The Marrakech benchmark contains 196 cultural POIs across ten heritage categories. Twenty calibration configurations were used to select the utility weights, and forty held-out paired configurations were used for the final comparison. On the held-out set, Agentic SHIFT achieved mean heritage coverage of 0.757, digital under-exposure of 0.767, cultural diversity of 0.522, expected suitability of 0.631, and a mean walking distance of 1.973 km. Relative to popularity ranking, it increased heritage coverage by 0.047 and digital under-exposure by 0.448 while reducing walking distance by 0.375 km. Removing the operational gate reduced operational validity from 1.000 to 0.696. Two stored OSMnx routes confirmed pedestrian-network execution within a four-hour budget. A documentary audit of thirteen narrative claims found eleven directly supported claims and two claims retained with explicit qualification, while a detailed audit of four visual records confirmed identity, creator, licence, and publication-suitable resolution for all four audited images.

Journal on Computing and Cultural Heritage
Cadi Ayyad University (MA)
Recommender Systems and Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon