- The paper reveals that current rigid token usage rights are driven by billing policies rather than technical constraints, opening avenues for flexibility.
- The paper details a design space analysis decomposing transferability into five axes and outlines five archetypal token movement types.
- The paper discusses how enabling token transferability can reduce user churn and foster market participation by reallocating usage rights across users, time, and services.
Transferability of Token Usage Rights in Generative AI Services: A Design Space Analysis
Introduction
The proliferation of generative AI systems has resulted in the widespread adoption of tokens as both the fundamental unit of language processing and a central economic instrument for metering user access. Current service models—exemplified by ChatGPT, Claude, Gemini, and Grok—utilize token-based billing structures that bind usage rights to individual accounts, specific platforms, and fixed subscription timeframes. The fixedness of these rights, however, is not an inherent technical constraint but a product of service-provider billing policy. This paper investigates the concept of token usage-right transferability as a design property and systematically analyzes design dimensions and the accompanying user and provider implications (2604.26683).
Structural Frictions in Current Token Usage Models
Token usage rights under prevailing AI service architectures embody rigid constraints along three primary axes: time, user (account), and service. Prepaid, subscription, or credit-based billing schemes create allocation-expiry mismatches, resulting in unused rights and direct user churn risk. The inability to reallocate or transfer token usage rights leads to inefficiencies and negative user experiences, substantiated by prior findings that expiring entitlements and service complexity drive churn and psychological resistance [breugelmans2017, cho2023].
While this rigidity can serve platform interests (e.g., preventing unauthorized sharing and driving breakage revenues), it imposes significant constraints on user autonomy and workflow flexibility. Practically, surplus and shortage co-occur across real-world teams and usage timelines, with no available mechanism to dynamically resolve them across accounts or service boundaries.
Design Space Analysis: Axes and Transferability Types
The paper leverages the Design Space Analysis (DSA) methodology of MacLean et al. (1991) to decompose transferability into five operational axes:
- Target: The specific resource moved (unspent usage, subscription rights, or personalized settings).
- Direction: The dimension along which movement occurs (time, user, service).
- Unit: The aggregation level of rights management (individual, team, organization, open market).
- Control: Assignment of decision power to users, platforms, or third parties.
- Reversibility: Degree to which transfers can be reclaimed or revoked.
Current commercial models group almost exclusively at the most restrictive ends: targets are immobile, directionality is blocked, only individual or organizational units are recognized, movement is platform-controlled, and no reversibility is relevant as no transfer is permitted.
The design space for transferability is then instantiated by five archetypal movement types:
- Carry-Over: Temporal extension of usage entitlements; analogous to telecom data rollovers.
- Co-Management: Resource pools shared within a defined user group (e.g., family plans).
- Transfer: One-to-one gifting or reallocation (paralleling gift cards).
- Conversion: Interoperability across different services/platforms (similar to airline mile alliances).
- Trade: Token marketplaces, supporting open reallocation (e.g., cloud reserved instance markets).
Each mode introduces specific tradeoffs regarding user experience, administrative complexity, market integrity, and the risk of abuse.
Implications and Prospects
The reframing of tokens—as objects not merely of economic or technical necessity but as design primitives—catalyzes several implications. Enabling transferability can mitigate churn drivers, expand market participation (by attracting risk-averse or cost-sensitive users), and foster new forms of user collaboration and autonomy. However, realizing such features demands substantial revision of both commercial terms and the supporting technical and regulatory infrastructures. Institutional control, enforcement, and dispute resolution mechanisms would need to be reconsidered, especially in high-reversibility or market-enabled scenarios.
Future work should empirically validate user and provider valuation of these axes and movement types, address organizational and legal feasibility, and model the complex game-theoretic dynamics arising from more liquid or open token economies.
Conclusion
This paper articulates a structured design space for the transferability of token usage rights in generative AI services, demonstrating that the current rigid binding of rights is a result of policy, not technical constraint. By moving tokens to the center of user-centered system design, the analysis offers an initial typology and framework for expanding user autonomy and market options. Realizing this vision, however, will depend on both technical and institutional developments, and on empirical validation of user and provider incentives in less restrictive models (2604.26683).