Why humans keep swiping left on AI

Technology promises growth, yet everyone has different perspectives on progress, and in that gap, suspicions take root

Featured faculty: Choi Sungwoo
Written by Putro Harnowo

While AI offers undeniable efficiencies, its integration carries unprecedented risks. The recent Pew Research Centre report shows that almost half of US adults use AI chatbots, but 40 per cent view AI as having a negative impact on society. The centre’s survey last year also found that more than one-third of respondents across 25 countries are concerned about AI.

“Many practitioners tend to focus on the technical benefits of AI, such as cost efficiency, speed, and convenience, but AI implementation often comes with unintended consequences, both positive and negative, for consumer behaviour,” says Choi Sungwoo, Assistant Professor of the School of Hotel and Tourism Management at the Chinese University of Hong Kong (CUHK) Business School.

In his co-authored study, Unintended negative consequences of AI in hospitality and tourism: A comparative bibliometric analysis and systematic review across three disciplines, Professor Choi, Kim Hyunsu of the University of Macau, and Hailey Shin of The Hong Kong Polytechnic University examine 233 research papers from psychology, business, and hospitality and tourism fields.

Across all three disciplines, Professor Choi highlights that these unintended consequences stem from human perceptions, with consumer backlash against humanlike AI, unpredictable impacts of AI on consumer decision-making, and employee friction with AI systems being the most critical.

These micro-level beliefs can scale into major problems, such as resistance or underuse of AI, service breakdowns, and even reputational risks. “Understanding these impacts can help practitioners and managers fine-tune their AI applications, reduce potential risks, and maximise the value of these technologies,” Professor Choi adds.

“Before implementing AI, managers should carefully consider how AI will affect both consumers and employees. They should put themselves in the shoes of the people who will interact with or work alongside the technology and then deploy AI in a way that enhances the overall service experience.”

The psychological reasons for AI resistance

Psychology has the longest timeline, with 74 papers on the unintended consequences of AI from 2012 to 2025, and it has laid the groundwork by investigating the individual-level mechanisms of human responses to AI.

For instance, mind perception theory argues that humans judge other entities by attributing two perceived mental capacities: cognitive functions and emotional functions. Machines like AI are perceived as having moderate cognitive but limited emotional capacities, suggesting they may be smart but lack the ability to feel or sense.

Given the tendency for some to distrust algorithms, even when they outperform human judgement, the term algorithm aversion was coined. It is a psychological bias in which people prefer human judgment and interaction, believing that algorithms lack human qualities like empathy or transparency and that humans understand situations better.

“Many practitioners tend to focus on the technical benefits of AI, such as cost efficiency, speed, and convenience, but AI implementation often comes with unintended consequences, both positive and negative, for consumer behaviour.”

Professor Choi Sungwoo

Additionally, since its underlying processes are too complex for laypeople to grasp, AI has faced criticism for its lack of transparency. To address this concern, scientists have developed tools and methods that enable humans to understand how an AI model makes its choices, known as explainable AI.

At the same time, algorithm appreciation, or the tendency for people to trust and rely on AI, has also gained attention. This occurs when AI meets both functional needs by effectively helping users with a task, and psychological needs by making users feel understood during the interaction. Missing either one can result in algorithm aversion.

Some consumers may display a strong machine heuristic, a mental shortcut in which they judge computer-generated work as truthful and accurate. While it sounds good, AI can hallucinate and is not immune to bias. Studies find that trust in AI depends on how human users perceive their own ability. If a person has less confidence in their own abilities, they are less likely to rely on AI.

Unintended consequences of AI in business applications

The business discipline has produced the most research on AI, with 108 papers from 2019 to 2024 in consumer-facing contexts. These studies emphasise rich, contextualised outcomes and their implications for organisations, employees, and broader society.

For instance, when an AI system fails, people tend to believe that other AI systems or AI in general will also fail. This phenomenon, called algorithmic transference, occurs because laypeople often see AI systems as indistinguishable from one another, so one failure is perceived to spread to other AIs.

Modern issues, such as technostress, or stress caused by technology, have become more important. Although technology makes life easier, it also introduces discomfort, such as techno-invasion, when technology feels like it is intruding on personal life, and techno-complexity, when technology is too difficult to understand. These anxieties make people less willing to embrace technologies.

In human resources, AI systems meant to help employees may backfire. People trust algorithms less for tasks perceived as subjective, such as those requiring human judgement, creativity, and emotional touch.

Social setting is also important. Consumers prefer robots to humans when interacting in awkward situations and feel less judged when communicating with AI. However, adding humanlike features may increase feelings of being judged due to perceived social presence.

Double-edged swords of AI in the service industry

Rather than treating AI as a generic technology, 51 studies spanning 2020 to 2025 in hospitality and tourism emphasise circumstances that influence how consumers perceive AI. Most influential studies acknowledge a persistent gap between technological sophistication and consumers’ feelings.

Studies on robot design reveal that a humanlike appearance can initially foster engagement but trigger discomfort when it becomes too human. However, incorporating humour, adjusting interaction timing, and engaging emotions or social functions may ease that feeling.

Service robots may also be perceived as a threat. The concept of AI identity threat refers to perceived psychological and sociocultural risks to human value and uniqueness. This threat varies between individuals and encounters. When consumers experience such threats, they fear that robots may steal or seize their welfare and identity.

On the other hand, privacy concerns raise awareness that these smart systems, whether service robots or online algorithms, constantly collect personal data. The main fear stems from how the data is gathered, accessed, and used, and from losing control over their digital footprint.

Algorithmic pricing is a niche theme covering airlines, hotels, and online travel agencies. While pricing algorithms help companies optimise prices and maximise revenue, consumers often view price discrimination as unfair and are keen to complain about it publicly.

Generative AI is expected to significantly impact trip planning, but studies show that it can worsen choice overload in travel decision-making. When AI chatbots help travellers narrow down options, travellers’ satisfaction with recommendations and intention to visit also tend to decline.