CARDLG in the AI/LLM era: continuities, accelerations and ruptures
| CARDLG dimension | Continuities (what persists) | Accelerations (what intensifies) | Ruptures (what Breaks from prior patterns) |
|---|---|---|---|
| Collections (C) | Libraries already collect tools alongside content (databases, discovery systems) | AI platforms compete with content budgets; tool collection becomes more central and expensive | Models contain hidden corpora libraries cannot audit; traditional collection ethics (known content) no longer apply |
| Access (A) | Licensing and authentication are familiar infrastructural concerns | Usage-based pricing and vendor lock-in intensify; access inequities may widen | User inputs/outputs (prompts, uploads) become data assets requiring governance – queries are now “content” |
| Retrieval (R) | Opaque ranking algorithms not new (discovery layers, vendor indexes) | Conversational interfaces intensify opacity; users see single answers not item lists | Retrieval becomes generative – systems produce new text rather than pointing to existing resources |
| Description (D) | Metadata work has always involved human judgment and embedded biases | AI can scale description work, reducing backlogs and increasing coverage | Probabilistic outputs are nondeterministic; same input yields different results, undermining systematic error correction |
| Literacy (L) | Critical evaluation of sources and systems has long been central to IL | AI increases volume of plausible misinformation; users rely on summaries over sources | Users must evaluate system trustworthiness, not just source credibility – requires understanding training data, uncertainty, bias |
| Governance (G) | Libraries have always navigated copyright, privacy, vendor negotiations | Platform power asymmetry intensifies; vendors are now technology giants with misaligned values | Platform decisions about training data and algorithmic behavior occur at scales/speeds libraries cannot control; governance becomes reactive |
| Cross-cutting tensions | Libraries balance openness, innovation and stability | Pressure to adopt tools users expect conflicts with deliberative risk assessment | Proprietary closed systems conflict with library values of openness; innovation speed outpaces governance capacity |
| CARDLG dimension | Continuities (what persists) | Accelerations (what intensifies) | Ruptures (what Breaks from prior patterns) |
|---|---|---|---|
| Collections (C) | Libraries already collect tools alongside content (databases, discovery systems) | AI platforms compete with content budgets; tool collection becomes more central and expensive | Models contain hidden corpora libraries cannot audit; traditional collection ethics (known content) no longer apply |
| Access (A) | Licensing and authentication are familiar infrastructural concerns | Usage-based pricing and vendor lock-in intensify; access inequities may widen | User inputs/outputs (prompts, uploads) become data assets requiring governance – queries are now “content” |
| Retrieval (R) | Opaque ranking algorithms not new (discovery layers, vendor indexes) | Conversational interfaces intensify opacity; users see single answers not item lists | Retrieval becomes generative – systems produce new text rather than pointing to existing resources |
| Description (D) | Metadata work has always involved human judgment and embedded biases | AI can scale description work, reducing backlogs and increasing coverage | Probabilistic outputs are nondeterministic; same input yields different results, undermining systematic error correction |
| Literacy (L) | Critical evaluation of sources and systems has long been central to IL | AI increases volume of plausible misinformation; users rely on summaries over sources | Users must evaluate system trustworthiness, not just source credibility – requires understanding training data, uncertainty, bias |
| Governance (G) | Libraries have always navigated copyright, privacy, vendor negotiations | Platform power asymmetry intensifies; vendors are now technology giants with misaligned values | Platform decisions about training data and algorithmic behavior occur at scales/speeds libraries cannot control; governance becomes reactive |
| Cross-cutting tensions | Libraries balance openness, innovation and stability | Pressure to adopt tools users expect conflicts with deliberative risk assessment | Proprietary closed systems conflict with library values of openness; innovation speed outpaces governance capacity |
Sharing content requires targeting cookies to be enabled. Please update your cookie preferences to use this feature.