Reading Robots: Software Robots for Extending Adaptive Reading Environments

Adaptive reading systems modify presentation, navigation, or assistance according to reader needs, while recent AI-enhanced readers add document-aware explanations and interaction. Yet the reading process itself is usually controlled directly by the user or by automation embedded inside the reader. This paper proposes Reading Robots (RR), a software-robot layer that operates a reading environment through semantic reading operations rather than through direct manipulation of interface coordinates or document DOM structures. RR is developed as an extension of the Universal Adaptive Reader (UAR), which provides a source-preserving Generated Reading Environment. The current RR-Core ver014 maps button, keyboard, and restricted text commands to a common deterministic command set, invokes UAR through RR Robot API ver001, manages reading jobs such as sequential read-all, and observes events and reading state. A central architectural rule is separation of ownership: RR controls what and when to read, whereas UAR retains ownership of speech text, pronunciation, text-to-speech, highlighting, and presentation. Human Override gives direct human actions priority over robot continuation. A 27-item API-conformance framework and trace logging support engineering verification. The contribution is an architectural one: reading is treated not only as an adaptive presentation activity but also as an executable, controllable, and verifiable job over semantic Reading Units.