Attention and selection
When AI can produce more ideas than people can examine, how should scarce human attention be used? We study ways to choose what deserves deeper evaluation without quietly excluding unusual or uncertain possibilities.
Research
What becomes scarce when knowledge becomes abundant?
Agalmic Research studies the gap between having more ideas and being able to do something worthwhile with them. The central questions concern attention, judgment, understanding, validation, representation and the practical use of knowledge.
Questions
Four connected areas.
These are research questions, not separate departments or claims that a new field has already been established.
Making knowledge useful
Why do good ideas fail to become useful knowledge or practical capability? We study the gaps between finding information, understanding it, testing it and being able to act on it.
Representation and discovery
How does the form in which information is presented change what people can notice, understand and discover? This includes experiments with visual and computational representations of complex data.
Economics of abundance
What changes when information, software, analysis or other valuable goods become much cheaper to produce? We study the new bottlenecks, scarce complements and shifts in value that follow.
Current study
How should scarce attention be allocated among too many plausible ideas?
Attention Allocation Under Cognitive Abundance is the main empirical study currently underway.
The first historical study successfully reconstructed 1,815 usable ICLR paper candidates, but that sample was not large enough to answer the planned question with the required precision. The study stopped before looking at the outcome data.
A larger follow-up is currently limited by the availability of historical data that can be collected without introducing selection bias. That is a useful result in itself: the present obstacle is trustworthy data, not more analysis code.
A real-world testbed
Representation can change what people notice.
Nemosyne is an independent project that can be used to test specific questions about representation and discovery.
For example, different ways of arranging the same dataset may make patterns easier or harder to see. Those effects can be measured rather than assumed.
Nemosyne does not belong to Agalmic Research, and software experiments cannot answer every economic or institutional question raised here.
For readers who want the machinery
Methods and research records remain open.
The site also keeps the detailed methods, prior-art work, negative results and provenance records behind the research. They are useful for scrutiny and reuse, but they do not need to be the first thing every visitor sees.