Building a Localised Muslim AI Model: Contextualising Behavioural Insights for the Middle East

𝗕𝘂𝗶𝗹𝗱𝗶𝗻𝗴 𝗮 𝗟𝗼𝗰𝗮𝗹𝗶𝘀𝗲𝗱 𝗠𝘂𝘀𝗹𝗶𝗺 𝗔𝗜 𝗠𝗼𝗱𝗲𝗹: 𝗖𝗼𝗻𝘁𝗲𝘅𝘁𝘂𝗮𝗹𝗶𝘀𝗶𝗻𝗴 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝘂𝗿𝗮𝗹 𝗜𝗻𝘀𝗶𝗴𝗵𝘁𝘀 𝗳𝗼𝗿 𝘁𝗵𝗲 𝗠𝗶𝗱𝗱𝗹𝗲 𝗘𝗮𝘀𝘁

Artificial Intelligence systems are only as good as the data they're trained on — and when it comes to people behaviour in the Muslim-majority regions of the Middle East, a generic model simply isn’t enough. To truly understand, serve, and optimise services in this context, we need to build Muslim-localised AI models that reflect the cultural, religious, and operational nuances of these societies.

𝗪𝗵𝘆 𝗟𝗼𝗰𝗮𝗹𝗶𝘀𝗮𝘁𝗶𝗼𝗻 𝗠𝗮𝘁𝘁𝗲𝗿𝘀 𝗶𝗻 𝗔𝗜

AI models trained on generalised global datasets often fail to detect or interpret cultural-specific behaviours. In the Middle East, where daily routines, crowd dynamics, retail patterns, and even gender segmentation differ significantly, a lack of localisation leads to:

• Misidentification of group vs. family behaviour

• Inaccurate assumptions on peak hours during Ramadan, Friday prayers, or Eid

• Blind spots in gender-segregated spaces (e.g. mosques, universities, shopping zones)

• Underrepresentation of modesty norms, such as full-body garments affecting AI vision systems

Without localisation, AI doesn’t just underperform — it misunderstands the people it’s meant to serve.

𝗨𝘀𝗲 𝗖𝗮𝘀𝗲: 𝗠𝘂𝘀𝗹𝗶𝗺 𝗕𝗲𝗵𝗮𝘃𝗶𝗼𝘂𝗿 𝗠𝗼𝗱𝗲𝗹𝗹𝗶𝗻𝗴 𝗶𝗻 𝗣𝘂𝗯𝗹𝗶𝗰 𝗦𝗽𝗮𝗰𝗲𝘀

Let’s take a mosque in Saudi Arabia or a shopping mall in Kuwait as examples. A localised Muslim AI model should be trained to understand:

• 𝗣𝗿𝗮𝘆𝗲𝗿 𝗧𝗶𝗺𝗲 𝗠𝗼𝘃𝗲𝗺𝗲𝗻𝘁 𝗣𝗮𝘁𝘁𝗲𝗿𝗻𝘀: Surge before and after salah times, different crowd flows on Fridays

• 𝗔𝗯𝗹𝘂𝘁𝗶𝗼𝗻 𝗥𝗼𝗼𝗺 𝗨𝘀𝗮𝗴𝗲: Special zones and higher movement intensity pre-prayer

• 𝗚𝗲𝗻𝗱𝗲𝗿-𝗕𝗮𝘀𝗲𝗱 𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗶𝗼𝗻: Respecting and modelling different flow patterns for men and women

• 𝗥𝗮𝗺𝗮𝗱𝗮𝗻-𝘀𝗽𝗲𝗰𝗶𝗳𝗶𝗰 𝗧𝗿𝗲𝗻𝗱𝘀: Late-night footfall, Iftar crowd surges, change in shopper engagement

A general AI model won’t pick up on these — but a localised Muslim AI model, trained with real-world Middle Eastern datasets, can.

What Does It Take to Build a Muslim AI Model?

𝟭. 𝗚𝗿𝗼𝘂𝗻𝗱 𝗧𝗿𝘂𝘁𝗵 𝗟𝗮𝗯𝗲𝗹𝗹𝗶𝗻𝗴 𝗯𝘆 𝗟𝗼𝗰𝗮𝗹 𝗔𝗻𝗻𝗼𝘁𝗮𝘁𝗼𝗿𝘀

Human-in-the-loop validation with Middle Eastern context is essential. Annotators must be aware of religious sensitivities, dress codes (abaya, hijab, thobe), and cultural norms.

𝟮. 𝗗𝗮𝘁𝗮𝘀𝗲𝘁 𝗗𝗶𝘃𝗲𝗿𝘀𝗶𝘁𝘆: 𝗚𝗲𝗻𝗱𝗲𝗿, 𝗘𝘁𝗵𝗻𝗶𝗰𝗶𝘁𝘆, 𝗔𝘁𝘁𝗶𝗿𝗲

Train AI vision to recognise people fairly — regardless of black abaya, white thobe, or face coverings. Avoid bias from underrepresented attire types in global datasets.

𝟯. 𝗖𝗮𝗹𝗲𝗻𝗱𝗮𝗿-𝗔𝘄𝗮𝗿𝗲 𝗜𝗻𝘁𝗲𝗹𝗹𝗶𝗴𝗲𝗻𝗰𝗲

Incorporate Islamic calendar events (e.g. Ramadan, Hajj season) into predictive models to better forecast footfall and usage patterns.

𝟰. 𝗣𝗼𝗹𝗶𝗰𝘆-𝗔𝘄𝗮𝗿𝗲 𝗧𝗿𝗮𝗰𝗸𝗶𝗻𝗴 𝗠𝗼𝗱𝗲𝘀

Some venues restrict full video recording due to privacy or religious reasons. AI must operate in privacy mode, counting without storing identifiable visuals.

𝗥𝗲𝗮𝗹-𝗪𝗼𝗿𝗹𝗱 𝗔𝗽𝗽𝗹𝗶𝗰𝗮𝘁𝗶𝗼𝗻𝘀: 𝗪𝗵𝗮𝘁 𝗖𝗮𝗻 𝗠𝘂𝘀𝗹𝗶𝗺 𝗔𝗜 𝗘𝗻𝗮𝗯𝗹𝗲?

• Smart Mosques: Zone occupancy, ablution area tracking, Friday crowd management

• Retail Analytics: Gender-specific aisle dwell time, campaign targeting for Eid or Ramadan

• Transport Hubs: Prayer room usage, adjusting crowd flow around Maghrib/Isha times

• Public Policy: Urban planning aligned with religious congregation patterns

A Step Toward Ethical, Contextual AI

At FootfallCam, our vision for Muslim-localised AI is part of a broader belief: context-aware AI is ethical AI. It's not just about building better models — it’s about building the right ones for the communities we serve.

Read the article here: https://www.footfallcam.com/BlogPost/Post/building-a-localised-muslim-ai-model-contextualising-behavioural-insights-for-the-middle-east

#footfallcam #peoplecounter #peoplecounting #ai #middleeast #muslimai #retailanalytics #smartmosque #privacy #gdpr #dataIntegrity #ethicalai