Découvrir Pandipedia

The applicable medical device software lifecycle-process standard is designated IEC 62304, formally titled “Medical device software: Software life cycle processes.”^[1]

Designing Effective Human-in-the-Loop Systems for High-Stakes AI

Human-in-the-loop design is not simply putting a person somewhere in an automated process. It means assigning decision authority deliberately, giving people enough information and time to act, monitoring the system throughout its lifecycle, and making a named person or institution accountable when things go wrong. Across aviation, finance, and health, the strongest common approach is risk-proportionate: the greater the system’s authority, uncertainty, and potential harm, the stronger the oversight, intervention, monitoring, and audit requirements should be.

The comparison below synthesizes EASA aviation guidance, Financial Stability Board sound practices for finance, and WHO guidance for health. These materials are not identical in legal status: EASA’s materials are emerging guidance, the FSB describes proposed proportionate sound practices rather than a binding international standard, and WHO’s materials are governance guidance. Their shared design logic is nevertheless practical for building high-stakes systems.

1. Set the oversight threshold before deployment

An oversight threshold is the point at which the AI must defer, request confirmation, trigger an alert, or stop acting. It should be determined from the consequence of error, the system’s operational authority, uncertainty, explainability, and the feasibility of human intervention, rather than from model accuracy alone.

A practical threshold policy should therefore define at least four states: assist, where AI informs a human decision; confirm, where a human must approve an action; override, where AI can act within limits but a human can intervene; and stop or escalate, where the system cannot safely continue. EASA’s framework explicitly distinguishes close oversight from remote or event-triggered intervention, while the finance and health guidance supports proportionality and clearly bounded tasks rather than a single universal control pattern.^{[12][13][14][15]}

2. Design the human interface around action, not presence

A nominal human reviewer is ineffective if the interface does not support comprehension, timely intervention, or meaningful challenge. EASA calls for information that is understandable, reliable, relevant, sufficiently detailed, and delivered at the right time, with design that supports situational awareness and cooperation.^[16][17] WHO similarly calls for explanations adapted to developers, professionals, patients, users, and regulators, while warning that AI should not substitute for professional or patient judgment.^[18][19]

The interface should also preserve human attention. Continuous supervision is appropriate only when the person can realistically monitor and intervene. Where continuous attention is not feasible, the system needs explicit operating constraints, reliable runtime monitoring, and alerts that identify when human intervention is required. EASA describes this distinction for remote oversight in aviation; the same design principle can inform other sectors, but sector-specific validation is still necessary.^[28]

3. Build a closed feedback and escalation loop

High-stakes oversight must continue after deployment. Approval or launch is not evidence that a system will remain safe as data, users, markets, clinical practice, or operating conditions change. EASA calls for continuous safety assessment using operational data and occurrences, the FSB calls for continuous monitoring and organizational learning, and WHO calls for quality control and improvement in real-world use.^[29][30][31]

4. Make accountability and auditability operational

Accountability requires more than naming an operator. Organisations should be able to show who approved the use case, who defined its limits, who validated it, who monitors it, who can suspend it, and who investigates harm. The FSB calls for documented lifecycle governance, designated senior oversight, and the ability to audit, explain, justify, and take responsibility for AI decisions and consequences.^[42][43][44]

Aviation guidance similarly places responsibility on the applicant or organisation to classify AI functions, justify the automation level, demonstrate that safety can be maintained throughout the product’s life, and retain records for monitoring and investigation.^[45][46] In health, responsibility remains with human and institutional stakeholders, governments retain responsibility for appropriate standards and oversight, and affected people should have routes to challenge decisions and obtain redress.^[47][48][49]

5. A practical best-practice framework

The following framework converts the cross-domain evidence into an implementation sequence. It should be adapted to the applicable regulator, professional standard, and use case rather than treated as a substitute for sector-specific approval.

Conclusion

The most reliable human-in-the-loop system is a controlled partnership, not a human approval button. Oversight should become stronger as authority, uncertainty, irreversibility, and potential impact increase. Interfaces should help people understand and act, feedback loops should detect and correct failures after deployment, and accountability should remain with identifiable human and institutional actors. The available guidance leaves some operational details open, including universal numerical thresholds, staffing levels, response times, and emergency workflows. Organisations therefore need to define and validate those details for each use case, document the rationale, and revisit them as real-world evidence accumulates.

The Economics That Will Shape Space Tourism Through 2036

Space tourism will probably remain a premium, capacity-constrained industry through the next decade, with suborbital flights developing before dependable orbital hospitality. The decisive economics are not launch prices alone: operators must combine reusable vehicles, high flight cadence, reliable suppliers, spaceport throughput, specialist insurance, passenger training, and enough affluent demand to spread fixed costs. The strongest near-term commercial opportunities are therefore likely to be logistics, research, training, station services, and ground infrastructure, with hotels in orbit emerging only if those foundations become reliable.^[1][2][3]

This report separates observed operating evidence from market forecasts. Published forecasts vary by more than an order of magnitude because some count only passenger flights while others include orbital stations, lunar missions, space hotels, or related services. A defensible planning range for a narrowly defined commercial passenger market is approximately USD 6 billion to USD 62 billion by 2035-2036, but even that range should be treated as scenario guidance rather than a precise prediction.^[4][5][6]

1. Supply chains, launch costs, and the importance of cadence

Reusable launch vehicles can reduce recurring hardware expenditure, but reuse becomes economically powerful only when vehicles fly often, refurbishment is predictable, and infrastructure can process missions repeatedly. Published Falcon 9 customer prices are approximately USD 62 million to USD 74 million, while an analyst estimate places the marginal cost of a reused-booster launch at roughly USD 15 million to USD 20 million. The difference is not operator profit alone: customer prices must also recover integration, operations, infrastructure, insurance, overhead, and development costs.^[7][8]

The evidence points to a learning curve rather than an immediate ticket-price collapse. A study of 4,405 orbital launches from 1960 to 2025 estimated that each doubling of cumulative payload delivered to orbit reduced average cost per kilogram by about 21.2%, from USD 87,023 per kilogram in 1960 to USD 3,868 in 2025, in 2024 dollars. Its projections reach roughly USD 1,600 per kilogram by 2030 and USD 273 to USD 300 by 2040, but these are system-level launch scenarios, not guaranteed costs for human-rated passenger vehicles.^[9][10]

Passenger operations will capture only part of these savings. Human-rating, abort systems, life support, medical screening, training, recovery, maintenance, certification, and passenger handling remain costly. NASA analysis indicates that reusable systems generally need high annual flight rates to amortise development and refurbishment, and that full reuse has a stronger economic case than partial reuse. The historical Space Shuttle illustrates the risk: it was designed for a two-week turnaround but achieved a shortest interval of 55 days, while its average flight cost was about USD 1.4 billion.^[11]

Manufacturing capacity could become the next bottleneck after vehicle designs mature. Space programmes depend on propulsion systems, composites, specialty alloys, additive-manufactured parts, thermal-vacuum and vibration testing, electromagnetic-compatibility testing, and spacecraft integration. Reported space-grade component lead times can exceed 18 months, while single-source suppliers create programme risk. Standardized components, common vehicle configurations, repeated booster use, and higher flight cadence are therefore central to price learning.^[12][13]

2. Insurance and risk allocation

Insurance is an operating input and a financing constraint. In the United States, licensed operators must demonstrate financial responsibility for covered third-party claims and specified risks involving government property and personnel. Required coverage is capped at the lesser of USD 500 million or the maximum liability insurance reasonably available worldwide for covered third-party claims, and at the lesser of USD 100 million or the maximum reasonably available for government-range property.^[20][21]

The model combines operator-funded insurance, reciprocal waivers, participant informed consent, and limited government protection for qualifying catastrophic claims above the operator’s required financial responsibility. Government indemnification can make extreme tail risk more financeable, but it is conditional rather than unlimited, and the retrieved statutory framework identifies September 30, 2028 as an important current endpoint for the relevant authority.^[22][23][24]

Passenger coverage is a separate issue from operator third-party liability. Specialist products may cover personal accident, medical and evacuation expenses, cancellation, baggage, passenger liability, or contingent life risks, while ordinary life insurance may exclude spaceflight or classify it as hazardous. Underwriters must assess age, health, acceleration, microgravity, life-support dependence, mission duration, vehicle reliability, and reentry conditions. The available evidence does not establish a reliable average passenger premium, so insurance should be treated as a material but currently unquantified addition to ticket cost.^[25][26]

The economic consequence is that informed consent does not make passenger risk inexpensive. Operators still need safety systems, certification, emergency procedures, and credible operating histories, while insurers and lenders face sparse loss data and potentially severe accident exposure. This favors established providers, repeatable vehicles, and jurisdictions whose liability rules make catastrophic risk sufficiently financeable.^[27][28][29]

3. Who will buy space travel?

The current customer base is small and affluent rather than mass-market. Virgin Galactic reported approximately 700 reservations and about USD 190 million in expected future revenue at a USD 600,000 base price, after previously selling roughly 100 tickets at USD 450,000. Blue Origin passengers have included business leaders, entrepreneurs, technical professionals, public figures, and sponsored participants; approximately 75 people had flown across 14 New Shepard missions, including five repeat passengers. These records show demand and some repeat participation, but they do not establish a broad repeat-purchase rate.^[30][31][32]

Orbital travel is a different market. Reported ISS mission prices were around USD 55 million per seat, with earlier Soyuz seats reported at USD 20 million to USD 35 million. Customers include ultra-high-net-worth individuals, mission sponsors, government-sponsored astronauts, researchers, and other specialist participants. Axiom missions connected with Saudi Arabia, Turkey, India, Hungary, and Poland, demonstrating institutional and national-representation demand beyond self-funded billionaires.^[33][34]

The clearest available willingness-to-pay evidence is old and should be treated cautiously: a 2002 Futron/Zogby survey of 450 affluent U.S. respondents found that 30% said they would pay USD 1 million for a two-week orbital trip, 16% USD 10 million, and 6% USD 25 million, while 70% were unwilling to pay any offered price. The survey indicates strong price sensitivity, but it is not a current market-clearing estimate and does not provide a formal space-tourism elasticity coefficient.^[35]

4. Market-size outlook and forecast discipline

Forecasts should be read as scenarios with different definitions, not as interchangeable measurements. Future Market Insights forecasts the market from USD 8.9 billion in 2026 to USD 62.1 billion in 2036, at a 21.8% CAGR, including suborbital, orbital, lunar expeditions, and space hotels. Its assumptions include reusable spacecraft, investment in spaceports and training, policy support, safety systems, and continuing premium demand; it reports suborbital at 58.4% of service type and short-duration missions at 71.3%.^[43]

Precedence Research provides a more conservative comparison: USD 1.36 billion in 2025, USD 1.58 billion in 2026, and approximately USD 5.94 billion in 2035, at a 15.88% CAGR. It covers suborbital and orbital travel, assumes that lower journey costs expand the customer base, and identifies reusable vehicles and U.S. market leadership, while acknowledging uncertainty around sustainability, price elasticity, and perceived value.^[44]

Other estimates range from USD 7.01 billion by 2035 from Spherical Insights to USD 53.7 billion from Global Insight Services. The Market Research Future endpoint of USD 88.6 trillion by 2035 is an extreme outlier. Its bottom-up framework reportedly multiplies available seats by ticket price and flight type, but the supplied evidence does not disclose the seat counts, ticket prices, flight cadence, or reconciliation needed to explain that result. It should not be treated as a consensus forecast.^[45][46][47]

5. Secondary industries: what scales first?

The most credible secondary industries are those with identifiable customers besides leisure travelers. Orbital logistics already has operational precedents in cargo delivery, orbital transfer, inspection, research services, station services, and satellite life extension. Contracts can cover resupply, refueling, repair, upgrades, and assembly, with economics improving when one vehicle serves multiple customers or extends the productive life of an expensive satellite. These activities are more likely to scale before mass orbital tourism because they can be supported by institutional and commercial users.^[51][52][53]

Commercial stations, research, and astronaut training are the next anchor markets. NASA seeks continuous human access to low Earth orbit and views commercial stations as training environments and proving grounds for lunar and Mars missions. Starlab is described primarily as a science and research platform that could expand research capacity while reducing the cost and time of sending experiments to orbit. NASA’s role as an anchor customer may help stations reach initial utilization, but the cited evidence does not establish station capacity, occupancy, or training throughput.^[54][55]

Spaceports and ground-side services should benefit from higher launch and crew-transport activity. Revenue mechanisms include launch and landing fees, vehicle processing, payload integration, propellant handling, maintenance, passenger handling, training facilities, and operator leases. Their economics depend on throughput because fixed facilities, regulatory staff, emergency systems, and transport links become more productive as launches and customers increase. Established, high-throughput sites are therefore more likely to be viable than isolated low-cadence facilities.^[56]

Conclusion: a staged commercial pathway

The likely sequence is suborbital travel first, followed by logistics and satellite servicing, then research and training, station services, and only conditionally orbital hospitality. This ordering follows the economics of customer breadth and utilization: logistics and institutional services can generate revenue from multiple customer groups, while orbital hotels depend on expensive stations being safe, certified, regularly supplied, and occupied by enough wealthy visitors.^[64][65]

By 2036, a market measured in several billion dollars is more defensible than a trillion-dollar interpretation, while a higher tens-of-billions outcome is possible if reusable vehicles achieve high cadence and the broader station and lunar categories are included. The central constraint will be capacity, not simply desire: launch slots, manufacturing lead times, spaceport throughput, certification, insurance, training, and maintenance must all scale together. Lower launch prices will help, but they will not by themselves make space travel mass tourism.^{[66][67][68][69][70]}