When AI Decides Who Belongs

Professor Fevzi Okumus of the University of South Carolina delivered a keynote at ITB Berlin on the equity and fairness risks embedded in generative AI as it transitions from experimentation to full deployment in travel and hospitality. He opened with a live audience poll revealing that most attendees use AI apps four or more times daily, contrasting this with the reality that just five to ten years ago these tools did not exist. Unlike Google, which returned multiple options for users to evaluate independently, today's AI synthesizes and ranks information directly—saving time but introducing inherited biases and errors that users rarely question.

Okumus drew on his own academic expertise to illustrate AI's unreliability: when he queried leading AI systems about the most prolific hospitality and tourism professors—an area where he has authored textbooks—the answers were only 50-60% correct, and descriptions of cities he personally lived in and wrote about contained notable gaps. He used the tangible example of searching for the best döner kebab restaurant in Berlin to show how AI recommendations may favor paid or chain operators over authentic local businesses, mirroring the pay-to-rank dynamics already seen in Google Search.

The core research concern centers on how large-scale data capture and predictive profiling create systemic disadvantages for marginalized travelers. AI systems build behavioral profiles from every search and interaction, enabling businesses to anticipate consumer decisions before they are made—a phenomenon Okumus described as companies being 'ahead of you before you make any decision.' This profiling can result in certain products, services, or travel experiences never being surfaced to specific demographic groups based on age, race, or gender, creating what he called 'invisible digital barriers beyond your control.'

Okumus also raised the hiring dimension: many companies now use AI to screen job applicants, and algorithmic filtering based on proxy variables can immediately exclude candidates from marginalized groups. On dynamic pricing, he noted that AI-powered revenue management tools continuously adjust prices based on situational signals, which can price out lower-income travelers and deepen economic exclusion.

Data consent and privacy erosion received significant attention. Okumus observed that nearly no one in the audience had read consent agreements line by line when signing up for AI apps—including himself—meaning users routinely grant unknown data-sharing permissions. He raised concerns about what would happen if personal AI query histories became public or were shared commercially, noting that companies sell aggregated profiles including email addresses, purchasing tendencies, and behavioral patterns, with enforcement varying dramatically by jurisdiction (stronger in Europe, absent in some countries).

A deeper cultural critique emerged around hospitality's core values. Okumus argued that over-reliance on AI-curated itineraries strips travel of serendipitous discovery—the chance encounter with an undiscovered café or local bakery—and risks homogenizing guest experiences into algorithmically standardized patterns. He expressed concern that the warmth of hospitality traditions (Turkish tea ceremonies, Japanese tea ceremonies, local greetings) could be reduced to AI-simulated replicas.

His forward-looking recommendations included: auditing AI training datasets for inclusivity and bias; establishing third-party validation frameworks and transparency standards that apply across jurisdictions; involving human domain experts to review AI recommendations in areas affecting marginalized travelers; creating industry-academic panels to assess AI's impact on equity; and co-creating ethical standards where profit is balanced against serving all communities. He predicted that by 2030, AI could surpass many humans across cognitive domains, with tour guides, travel agents, and even university professors facing genuine displacement.