Graceful forgetting: Memory as a process
This work addresses the fundamental problem of memory capacity for cognitive science and AI, offering a new theoretical framework that is foundational but incremental in building on existing knowledge.
The authors tackled the problem of explaining how unbounded sensory input fits into bounded memory by proposing a framework where memory is stored as statistics that are repeatedly summarized and compressed, guided by heuristics to optimize for future needs. The result is a novel account emphasizing memory as an intensive, complex process using statistical representations.
A rational framework is proposed to explain how we accommodate unbounded sensory input within bounded memory. Memory is stored as statistics organized into structures that are repeatedly summarized and compressed to make room for new input. Repeated summarization requires an intensive ongoing process guided by heuristics that help optimize the memory for future needs. Sensory input is rapidly encoded as simple statistics that are progressively elaborated into more abstract constructs. This framework differs from previous accounts of memory by its emphasis on a process that is intensive, complex, and expensive, its reliance on statistics as a representation of memory, and the use of heuristics to guide the choice of statistics at each summarization step. The framework is intended as an aid to make sense of our extensive knowledge of memory, and bring us closer to an understanding of memory in functional and mechanistic terms.