The research team set out to understand what actually happens when people turn to tools such as conversational AI assistants while organising a real trip. Rather than running a laboratory experiment, the researchers drew on interviews with a group of UK adults who had used generative AI to help plan a recent or upcoming journey. From these accounts, they built a framework explaining how the technology redistributes mental work during the planning process. The central idea is subtle but important: generative AI does not simply automate holiday planning. Instead, it takes over specific coordination tasks, such as organising information, tracking details and keeping options ordered, which frees travellers to concentrate on the judgements that matter most to them.

This distinction between redistributing mental work and removing it runs through the whole study. The researchers describe a planning process built from a series of interconnected decisions about transport, accommodation, budgets and activities. Each decision constrains the others, and the sheer volume of information available online can quickly tip an engaged planner into a state of overwhelm. In that state, people tend to procrastinate, abandon their plans or settle for the first acceptable option simply to end the strain. The framework shows how an AI assistant can provide structure at exactly these pressure points, helping travellers stay focused and making it easier to move forward without hijacking the final choice.

According to the findings, generative AI supports travellers in three main ways. First, it reduces the mental effort of organising, tracking and coordinating plans, which in turn eases feelings of overwhelm. Second, it creates room to pause, compare options and reflect before committing to a decision, a form of breathing space that many interviewees found difficult to achieve when planning alone. Third, it helps travellers get started, keeps the planning process moving and carries it through to completion. For some participants, this third function was transformative: a holiday that might have been delayed indefinitely or abandoned entirely became something they felt able to finish planning.

Dr Yoonjung Kim, a doctoral researcher in Hospitality and Tourism Management at the University of Surrey and lead author of the study, explained the mechanics behind this effect. Planning a holiday, she noted, involves a series of connected decisions that can quickly become overwhelming, particularly when travellers face endless information and choices. The research finds that generative AI can provide structure, help people stay focused and make it easier to move forward without taking the final decision out of their hands. In cognitive terms, the technology acts as an external memory and organising layer, absorbing the clerical side of planning so that limited human attention can be spent on comparison and evaluation.

The benefits were most pronounced among travellers who found planning hardest to manage. The study suggests that AI turned holiday planning from a task they might delay or give up on into something they felt capable of completing. This is a significant finding because it reframes AI travel tools not merely as conveniences for tech-savvy users but as potential accessibility aids. The same properties that make generative AI useful for a busy professional comparing flight combinations, the ability to summarise options, hold context across a long conversation and keep multiple constraints visible, can make the difference between a possible holiday and an impossible one for someone who struggles with executive function or information overload.

Professor Iis Tussyadiah, Dean of Surrey Business School and co-author of the study, emphasised this broader implication. For some travellers, she observed, generative AI is more than a convenient shortcut: it can make a demanding decision-making process feel manageable and, in some cases, make planning feel possible. She argued that the findings show how travel technology could become more inclusive by being designed around the different ways people think, plan and make decisions. That framing points toward what the researchers call cognitively inclusive travel technology, systems that accommodate diverse cognitive styles rather than assuming a single idealised planner who processes information linearly and effortlessly.

The researchers highlight one group that could benefit substantially from such design thinking: neurodivergent travellers, for whom the open-ended, multi-constraint nature of trip planning can be especially taxing. At the same time, they are careful to note that cognitive overload is not a niche problem. Any traveller can hit a planning wall when a trip involves large amounts of information, multiple constraints and interconnected decisions. A family coordinating a multi-generational holiday, a traveller managing a medical requirement abroad, or simply anyone facing a saturated search-results page all encounter the same underlying difficulty: the working memory required to hold a trip in mind exceeds what is comfortably available.

Dr Bora Kim, Associate Professor in Surrey Hospitality and Tourism Management and co-author of the study, offered a forward-looking perspective on what this means for the travel industry. The real promise of AI in travel planning, she argued, is not simply speed or convenience. The same tool can support different travellers at different points of difficulty while allowing them to retain control over their decisions. She called for travel technologies designed to provide personalised cognitive and emotional support while protecting travellers’ autonomy, a formulation that treats AI as an assistant operating under human direction rather than an agent making purchases and itineraries on its own initiative.

The study’s conceptual framework arrives at a moment when generative AI is rapidly being woven into consumer travel platforms, from airline booking assistants to chatbot concierges. By identifying where AI support genuinely helps, reducing coordination burden, enabling reflection, and sustaining momentum through to completion, and where human control must remain, the Surrey research offers designers and travel companies a map of the terrain. It also offers travellers a reassuring message: the goal of AI in holiday planning is not to choose your trip for you, but to carry enough of the administrative weight that choosing becomes enjoyable again. For the overwhelmed majority who have abandoned a booking page at midnight, that distinction may prove to be the difference between another unrealised plan and a departure date on the calendar.

Subject of Research: Generative AI as cognitive scaffolding in travel decision-making and holiday planning

Article Title: AI could make holiday planning easier for overwhelmed travellers

Article References: AI could make holiday planning easier for overwhelmed travellers. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: generative AI, travel planning, decision-making, cognitive load, University of Surrey, Journal of Travel Research, tourism technology, neurodivergent travellers, cognitive scaffolding, consumer behaviour, AI assistants, travel industry

The intervention, known as the Hierarchical Questioning Strategy (HQS), is grounded in a deceptively straightforward idea: comprehension develops in stages, and readers must consolidate factual, literal understanding before they can reliably infer implicit meanings or critique an author’s argument. The researchers operationalized this through a Literal–Inferential–Evaluative (L-I-E) sequence. At the literal level, students retrieve explicitly stated information—facts, main ideas, sequences of events. At the inferential level, they combine textual evidence with background knowledge to read between the lines, identifying cause and effect, author tone, and unstated themes. At the evaluative level, they render critical judgments, take positions, and assess the strength of arguments. Each rung of the ladder is verified before the next is attempted, ensuring that no student is asked to analyze a text they have not yet decoded.

The theoretical architecture behind this sequence draws on several well-established frameworks. Schema theory, as articulated by cognitive scientist David Rumelhart, holds that readers use existing mental frameworks to decode explicit textual facts, forming the literal foundation of comprehension. Levels of Comprehension theory adds that mastering literal detail is a prerequisite for inferential and evaluative processing. The strategy also operationalizes Lev Vygotsky’s Zone of Proximal Development: the literal tier establishes a student’s independent baseline, the inferential tier introduces a guided challenge just beyond it, and the evaluative tier pushes learners toward the upper bound of their potential. When text difficulty outstrips a learner’s linguistic capacity, working memory becomes consumed by mechanical decoding, leaving no cognitive resources for higher-level processing—a barrier that drives frustration and disengagement. The HQS is designed precisely to prevent that overload.

The study’s context gives these findings particular urgency. English has been Ethiopia’s medium of instruction in secondary and higher education for more than a century, yet many Ethiopian EFL students remain at a surface-level grasp of texts. The researchers point to two systemic problems: local instructional materials often lack a balanced distribution of cognitive question levels, and universities rarely provide explicit comprehension training. Traditional Ethiopian reading instruction follows a pre-, while-, and post-reading framework, but its implementation leans heavily on skimming, scanning, and vocabulary work—surface-level skills—while inferential and evaluative comprehension is largely overlooked. The preparation gap is severe: national data cited in the study indicate that only 3.2 percent of high school seniors achieved passing marks on higher education entrance examinations, meaning incoming freshmen often arrive with low reading self-efficacy and significant literacy deficits.

To test whether structured questioning could close that gap, the research team, led by Endalfer Melese Moges with advisors Dagne Tiruneh Dinsa and Tewodros Zeleke Mendida, employed a non-equivalent control group quasi-experimental design. Participants were pre-medicine freshmen enrolled in a compulsory Communicative English Language Skills course. Two intact class sections were selected, and a lottery assigned one to the experimental condition (n = 53) and the other to the control condition (n = 54). The experimental group received eighteen fifty-minute sessions of explicit HQS instruction over four months, following a gradual release of responsibility model: the instructor first explained and modeled each questioning level, then guided collaborative practice, and finally stepped back for independent application. The control group followed the conventional module-based routine of background activation, silent reading, and standard comprehension questions without explicit cognitive scaffolding.

The researchers took extensive precautions to protect the validity of the comparison. The same instructor taught both conditions using scripted lesson plans, classes were scheduled on different days and shifts to block peer communication between groups, and intervention materials were password-protected and tracked. Fidelity was monitored through unannounced supervisor observations using a standardized checklist. Before the intervention, independent-samples t-tests confirmed the groups were statistically equivalent in both reading achievement and self-efficacy, with p-values well above the significance threshold. The instruments themselves were rigorously validated: a 35-item reading comprehension test spanning all three cognitive levels showed strong test–retest reliability (r = 0.89), open-ended items achieved high inter-rater agreement (Cohen’s kappa = 0.87), and a 16-item reading self-efficacy questionnaire adapted from Bandura’s framework yielded a Cronbach’s alpha of 0.846.

The post-intervention results were dramatic. The experimental group’s mean reading comprehension score rose to 11.87, compared with 9.50 for the control group, and the corresponding self-efficacy means were 4.27 versus 3.65 on a five-point scale. Independent-samples t-tests found highly significant differences favoring the experimental group on both outcomes (p < 0.001). A one-way MANOVA produced a Wilks’ Lambda of 0.224 with a partial eta squared of 0.776, meaning group membership accounted for 77.6 percent of the variance in the combined outcome measures—an exceptionally large multivariate effect. Follow-up univariate analyses showed the intervention explained 74 percent of the variance in reading scores and 41 percent of the variance in self-efficacy. The effect size for reading achievement, Cohen’s d = 1.67, sits far beyond the benchmarks typically used to define a large educational effect.

The psychological mechanism, the authors argue, is as important as the cognitive one. Drawing on Albert Bandura’s Social Cognitive Theory, they contend that mastery experiences—successive, incremental personal triumphs—are the most powerful catalyst for self-efficacy. As students progressed through the L-I-E stages, each verified success at one level built the confidence needed to attempt the next. The structured progression, they suggest, shifts a learner’s orientation from affective apprehension to cognitive agency: reading ceases to be a passive decoding exercise and becomes an active, goal-directed process. Notably, the intervention’s effect on actual achievement was more pronounced than its effect on self-perceived capability, which the researchers interpret as evidence that self-efficacy may adapt more gradually, partially mediated by historical academic self-concepts and broader external variables.

The findings carry clear implications for classrooms well beyond Ethiopia. Globally, university students show a steep proficiency drop when moving from literal to higher-order tasks—one cited analysis found success rates falling from 56 percent on literal reading to just 22 percent on evaluative tasks. The Jimma results suggest that this drop is not an inevitable feature of foreign-language reading but a consequence of instruction that demands evaluative output without first securing the literal baseline. The authors recommend that instructors enforce sequential mastery, ensuring students demonstrate solid factual understanding before introducing abstract interpretation; that teacher educators institutionalize training in hierarchical questioning through workshops and peer coaching; and that curriculum developers revise tertiary EFL materials to prioritize deep textual analysis over superficial comprehension checks.

The study’s limitations temper, but do not erase, its promise. The sample was small, non-randomized, and drawn from a single pre-medicine cohort at one university, and the four-month intervention cannot reveal whether gains persist once scaffolding is removed. Self-report bias in the efficacy measure and unmeasured differences in teacher expertise are additional caveats. The authors call for randomized controlled trials across multiple sites, longitudinal designs tracking retention, and qualitative triangulation through classroom observation and student interviews. Even so, the core message is likely to resonate widely: questioning, often treated as a tool for checking recall, can be reimagined as a dynamic instructional engine—one that simultaneously builds the critical reading skills students need for academic survival and the self-belief they need to keep reading when texts get hard.

Subject of Research: Effect of hierarchical questioning strategy on EFL reading comprehension and self-efficacy among Ethiopian university freshmen

Article Title: Improving freshman EFL reading comprehension and self-efficacy through hierarchical questioning

Article References: Moges, E. M., Dinsa, D. T., & Mendida, T. Z. (2026). Improving freshman EFL reading comprehension and self-efficacy through hierarchical questioning. Discover Education, 5(1), Article 1086. https://doi.org/10.1007/s44217-026-02192-4

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02192-4

Keywords: hierarchical questioning, reading comprehension, self-efficacy, EFL, higher education, Ethiopia, quasi-experimental design, cognitive scaffolding, Bloom's taxonomy, Zone of Proximal Development, language education, instructional design