Evidence (1286 claims)

Practical barriers—software access, available datasets, and lab time—limit experiential learning that builds AI competency.

Respondents cite limited opportunities for applied, project-based learning with AI tools; where AI appears in curricula, coverage is more theory-oriented than hands-on.

Many institutions lack clear, consistent, or context-sensitive policies for AI use in learning, assessment, and academic integrity.

Educators frequently report lower confidence in teaching AI-relevant skills than students report in using AI tools, reducing instructional capacity.

Rural digital divides mean AI benefits will be unevenly distributed; models trained on digitally-rich urban records could bias resource allocation away from rural trainees.

Key disadvantages and barriers to the proposed digital modernization are administrative backlogs, rural infrastructure deficits, and qualification fragmentation.

Rural constraints (limited electricity, limited ICT access, and fewer training centers) reduce inclusion of rural trainees in vocational-to-engineering pathways.

Fragmentation and overlap across vocational and technical qualifications create discontinuities that impede career progression.

Administrative irregularities and backlogs exist in SAQA/NATED ratification processes, including suspension or deregistration actions carried out without due process.

Misalignment between hands-on technical training (artisan-level skills) and formal institutional certification (SAQA/NATED/NCV/SETA) is blocking vocational-to-engineering career progression.

Trade policy (trade openness) should be modeled as a moderating factor when estimating technology-driven urban outcomes because openness can dampen local price effects of digital trade.

Greater trade openness weakens (attenuates) the positive effect of digital trade on city-level house prices.

Policy and regulatory vacuum (data governance, interoperability, safeguards) limits scale and inclusive diffusion of AI in agriculture.

Limited digital literacy and human capacity among smallholders is a key barrier to adoption and effective use of AI solutions.

Scalable adoption of AI in developing-country agriculture is constrained by infrastructure gaps (connectivity, power, data platforms).

Unequal access to high-quality AI tools creates demand-side market failures and vendor concentration risks, justifying public intervention (subsidies, procurement tied to privacy/audit requirements).

Traditional signals (test scores, credentials) may lose reliability as AI assistance becomes widespread, which will alter estimates of skill endowments and returns to education.

Teachers currently lack sufficient preparedness (training, time, tools) to integrate AI into formative assessment and to interpret AI-informed evidence; addressing this is necessary for successful transition.

Unequal access to AI amplifies existing achievement gaps and biases assessment outcomes, making equity a primary concern for AI-compatible assessment.

AI changes the production of student work (e.g., generative content, altered authorship), undermining traditional notions of student-authored artifacts used in assessment.

Standardized summative tests were designed for an environment without routine, external AI assistance; those design assumptions are breaking down.

Conventional standardized, summative assessment is becoming increasingly misaligned with classroom reality because widespread student access to AI tools changes what, how, and where learning occurs.

Students raised concerns about ChatGPT producing factual errors, the risk of overreliance that could reduce independent thinking, and functional constraints of free ChatGPT versions.

Biased or unrepresentative AI outputs produce negative externalities, including maladaptation and inefficient investments in vulnerable regions.

Returns to scale in compute and data favor incumbents; without intervention this dynamic can entrench inequality in the global climate-information market.

Concentration of compute and model development creates market power for Northern institutions and companies, likely leading to unequal pricing, control over standards, and capture of high-value climate services.

Rapid AI adoption without a shift from model-centric to data- and equity-centric development risks producing systematically worse performance and misleading recommendations for the most climate-vulnerable, data-sparse regions.

Large language models (LLMs) that rely on dominant, textualized climate knowledge tend to foreground Northern epistemologies and marginalize local or indigenous knowledge, reinforcing biases in climate narratives and recommendations.

In climate impact modelling, sparse and unrepresentative exposure and vulnerability data combined with inadequate validation generate high uncertainty and risk of misleading interventions and maladaptation in vulnerable locales.

In weather and climate modelling, historically and spatially biased observational data produce systematic performance gaps in under-observed tropical and low-income regions, reducing forecast fidelity where adaptive capacity is lowest.

The geographic concentration of compute and model development creates path dependence: model design, training datasets, and validation reflect Northern priorities and contexts.

Numerical simulations using calibrated parameter sets produce phase diagrams and time-paths that show when gradual adjustment transitions into explosive demand collapse and financial stress under different combinations of capability growth, diffusion speed, and reinstatement rate.

Because consumption is concentrated and top incomes have high AI exposure, shocks to top-income labor/income disproportionately affect aggregate consumption and thereby threaten private credit and mortgage markets — the paper maps plausible exposures to roughly $2.5 trillion of global private credit and about $13 trillion of mortgages.

Top-quintile households are also the cohort with the highest measured AI exposure (i.e., incomes/occupations most exposed to AI substitution), increasing the concentration of AI-driven demand risk.

Intermediation collapse: AI agents reduce information frictions and automate advice/coordination tasks, compressing intermediary margins toward logistics/execution costs and repricing business models across SaaS, payments, consulting, insurance, and financial advisory, with knock-on effects for firm valuations and collateral values that underpin credit markets.

Ghost GDP: AI output that replaces labor-intensive output can create a wedge between measured GDP (which may rise) and consumption-relevant income (which can fall) because a declining labor share reduces monetary velocity absent proportionate transfers — producing hidden demand shortfalls.

When firms rationally substitute AI for labor, aggregate labor income can fall and lower demand, which accelerates further AI substitution — a 'displacement spiral' whose net feedback is either self-limiting (convergent) or explosive (runaway adoption + demand collapse) depending on AI capability growth rate, diffusion speed across firms/sectors, and the reinstatement rate (rate at which new paid human roles or demand reappear).

Rapid AI adoption can create a macro-financial stress scenario not primarily through productivity collapse or existential risk but via a distribution-and-contract mismatch: AI-generated abundance reduces the need for human cognitive labor while institutions (wage contracts, credit, consumption patterns, financial intermediation) remain anchored to the scarcity of human cognition, producing a self-reinforcing downward spiral in labor income, demand, and intermediary margins that can tip into an explosive crisis unless offset by sufficiently fast reinstatement of human-paid demand or deliberate policy/market responses.

Introducing ‘agent capital’ (AI that lowers coordination costs) reduces coordination costs inside firms (coordination compression).

Extractive industries often deliver limited local employment and mainly generate rents rather than broad employment or skill spillovers.

FDI may increase within‑country wage inequality, especially when concentrated in extractive sectors or low‑skill activities.

FDI may deepen labor market dualism: creating formal, higher‑paying jobs for a minority while many remain in precarious, low‑pay informal work.

A one standard-deviation increase in AI adoption lowers wages in the middle income quintile by 1.4%.

Loss of control over research data impedes local capture of value (knowledge, IP, downstream services) and can create externalities when data are repurposed or commercialized without equitable benefit sharing.

Dominant AI/cloud providers become de facto gatekeepers of data processing and storage; researchers and institutions, particularly in lower‑capacity jurisdictions, have limited bargaining power to enforce data‑sovereignty or transparency terms.

Algorithmic opacity and cross‑border regulatory fragmentation raise monitoring, compliance, and contractual costs for collaborative research, effectively increasing the transaction costs of data‑intensive science.

Inequalities in infrastructure (local compute, storage, institutional procurement power) amplify these problems: researchers in weaker jurisdictions face higher risks and fewer mitigation options.

Rather than shifting liability away from researchers, AI systems increase researchers' ethical responsibilities: researchers must assess third‑party tools, negotiate data flows, and manage risks despite having limited contractual leverage.

Algorithmic opacity (hidden models, undocumented data flows, proprietary cloud stacks) reduces researchers' ability to control or even know how participant data are used, transferred, or monetized.

Everyday AI services used in research introduce new, diffuse points of data capture and processing that complicate informed consent and privacy management.