01
Prediction
Occupancy, arrivals, workload, and required labor forecasts.
Primary issue: comparative predictive capability.
Research agenda
How should decision authority be divided between hotel managers and intelligent systems?
Hotel Decision Lab studies how technology, data, analytics, automation, and artificial intelligence affect managerial decision-making inside hotel operations. The current application is AI-supported hotel labor decision-making.
The objective is not maximum trust, acceptance, automation, or manager control. It is calibrated reliance and better decision quality.
Current canonical research question
Under what conditions should hotel organizations pre-assign managers the authority to accept, modify within bounds, pause, override, or escalate an AI-supported labor recommendation, given the system's demonstrated comparative capability, data coverage and uncertainty, the manager's access to nonduplicative local information, decision stakes and reversibility, employee and guest consequences, error detectability, feedback speed, incentives, and accountability?
The recommendation instance remains the primary unit of analysis, but authority design occurs before the recommendation appears.
Decision architecture
Hotel labor is not one decision.
02
Resource allocation
Weekly schedules, departmental staffing, housekeeping workload, and overtime distribution.
Primary issues: objectives, constraints, employee impact, fairness, feasibility, and discretion.
03
Real-time personnel action
Sending someone home, calling someone in, extending a shift, reducing coverage, or moving labor between departments.
Primary issues: local context, consequences, reversibility, time pressure, and accountability.
Current research position
Human review is not inherently protective.
Its value depends on what the manager adds, which errors the manager can detect, and whether the organization enables useful intervention and learning.
Proposed condition 01
Informational advantage
The manager contributes timely, relevant, material information unavailable to or inadequately represented in the system.
Proposed condition 02
Diagnostic competence
The manager can recognize consequential system error and intervene effectively.
Proposed condition 03
Governance alignment
Authority, incentives, accountability, time, and escalation structures support useful intervention.
Status: These conditions are proposed concepts derived from existing research and hotel operating logic. They have not been empirically validated in hotel labor operations.
Evidence position
What the current review supports, and where it stops.
Stronger supported conclusions
Calibrated reliance, not agreement
Trust, acceptance, and agreement are inadequate on their own. People can both reject useful algorithms and over-rely on them. Modification affects adoption more reliably than decision quality, and required human approval is not automatically safer.
Mixed or conditional
Complementarity and explanation
Structured human-model combinations can outperform either source when the human contributes distinct information and recurring bias is controlled. Explanation helps only when it improves diagnosis rather than persuasion or perceived understanding.
Hotel-specific evidence
Relevant, but not decisive
Reviewed hotel studies address control preferences, employee responses to AI, decision-support literature, occupancy prediction, and staff-scheduling systems. They do not establish whether a manager action improved a live AI-supported labor decision.
Transferred evidence
Useful mechanisms, uncertain transfer
Much of the behavioral evidence comes from forecasting, human factors, online experiments, consumer tasks, retail, and organizational settings. It informs propositions, but does not become direct hotel evidence.
Within the supplied evidence set, no reviewed full-text study directly linked hotel managers' recommendation-level actions to downstream labor, service, employee, guest, or financial outcomes. This is a bounded search result, not a claim that no such research exists.
Decision rights
Design-time authority and run-time response are different.
Proposed design mode
Human-led
The manager owns the decision and uses system information as an input.
Proposed design mode
AI-recommended
The system recommends an action, but the manager decides.
Proposed design mode
AI-generated with required review
The system produces the intended action, but a defined review is required before implementation.
Proposed design mode
Automated with human monitoring
The system acts unless a threshold, exception, or later review triggers intervention.
These categories are proposed and unvalidated. At run time, available responses may include accept, modify within bounds, pause, investigate, override, or escalate.
Five priority research questions
What Hotel Decision Lab needs to learn next.
- Which manager-only facts measurably improve hotel labor decisions, and which recurring adjustments add bias or noise?
- For which hotel decisions does bounded modification preserve model value, and what evidence should allow a manager to exceed the bound?
- How should hotel companies define decision quality when labor cost, service, employee burden, guest consequences, fairness, and reversibility conflict?
- How do uncertainty displays, explanations, and visible failures affect calibrated reliance across seasons and property-specific regime changes?
- How do ex ante authority structures and labor pressure affect information gathering, acceptance, override, pause, escalation, accountability, and downstream outcomes?
Published work
Current flagship essays.
Research-informed conceptual essay
When Hotel Managers Override the Algorithm
Why human review should be evaluated through information, diagnostic competence, and governance rather than assumed to create safety simply because a manager remains involved.
Read the articleResearch-informed conceptual essay
When Local Knowledge Beats the Forecast
When manager-supplied local information deserves decision weight, when intervention adds noise, and how hotels can make those information claims visible and evaluable.
Read the articleFor the evidence process, access rules, transfer boundaries, and publication review gate, read the Hotel Decision Lab methodology.