SCHOLARS’ COUNCIL FOR AI
AI Research for Human Judgment and Life
A faculty-centered academic research organization that studies the conditions of judgment and responsibility needed when AI takes part in people’s lives and in the work of organizations.
At a glance
- Faculty who have agreed to participate 4
- Research areas 8
- Initial projects 2
- Social goals 4
Why this research
When work is handed to AI, what else is handed over with it?
The more we use technology that does our work for us, the more we also hand over authority that affects our information, our relationships and our choices. The Scholars’ Council for AI seeks to connect the knowledge that academic disciplines have built up to the work of designing and using AI, and to verify whether that knowledge actually helps people in real-world settings. Its starting point is the idea that people should be able to understand, choose and correct their own work even while receiving help from AI.
The change sought
The change the Scholars’ Council for AI seeks
Knowledge should reach those who need help, people who are unfamiliar with these tools should also gain chances to learn, and people should be able to demand an explanation and a correction when faced with a wrong decision. The change sought is for progress in AI to lead to these everyday possibilities.
Reducing harm
The right to judgment and accountability
Human capability
Gaps in access and use
Scope
What does the research cover?
There are eight research areas, and they overlap. Several fields can take part together in a single project.
Knowledge and Meaning
How are the sources of knowledge and the conditions for applying it maintained?
Reasoning and Judgment
On uncertain evidence, which conclusions are chosen and when is judgment withheld?
Action and Learning
How are the results of actions checked and reused?
Collaboration and Organization
How are roles and information divided and coordinated?
Humans and AI
How are judgment and authority shared?
Creativity and Innovation
How are problems and alternatives framed anew?
Trust and Responsibility
On what grounds are the scope of use and responsibility set?
Evaluation and Empirical Validation
How are actual effects, costs and failures checked?
Projects to examine first
What is taken up first?
DPIM Decision Process Intervention Model
- Observing the stateroles · evidence · unconfirmed items
- Choosing an interventionmore evidence · re-examination · alternativesno intervention is also included in the comparison
- Evaluating the effect of the interventionresults · time · cost · errors
A research agenda that seeks to learn where, and with what kind of intervention, results improve in the judgment process of an AI team. The comparison should include cases with no intervention.
KAC Knowledge Action Chain
- Knowledge and methods needed
- Evidence and conditions applied
- Execution records and outputs
- Confirmed results and limits
From confirmed results and limits back to the knowledge and methods needed — reviewed experience leaves conditions of application to be tested again
A research concept that deals with how knowledge leads to execution and to verified outcomes.
DPIM is examined under the hypothesis that it may contribute, in the first instance, to reducing harm caused by wrong judgments, and KAC under the hypothesis that it may contribute, in the first instance, to tracing evidence and execution results and to responsible operation. To say that the two projects also contribute to human capability or to gaps in use, a separate evaluation that includes people and the field is needed.
See the research that deals with both projectsSee the glossary
Dividing roles
In joint research, who does what?
Joint research can be run in a way where different roles carry forward a single process of verification.
The scholarly review of the Scholars’ Council and the participation of the Enterprise and the field come together in a joint research question, and the Research Institute can turn that question into a form that can be verified.
Scholars’ Council: The Scholars’ Council establishes research directions and criteria for judgment on the basis of accumulated scholarship.
Enterprise: The Enterprise implements research results as technologies, products and services, and operates them.
Research Institute: The Research Institute turns scholarship into concrete research models and carries out experiments and verification.
Faculty who have agreed to participate
Won-Woo ParkHyo-Gun KimNamgyoo K. ParkSeokhwa YunSee the facultyFrequently asked questions
How far can research results be trusted?
To weigh the trustworthiness of AI, one should first state specifically what it is that is being called trustworthy. The information needed for reproduction and review is kept as the record of each study. Publishing successful and failed cases together helps later research avoid repeating the same limits.
Does adopting it have an effect?
Not every difference between before and after a particular system is adopted is concluded to be an effect of that system. Comparison conditions are kept where possible, and differences that could not be kept are stated as limits. Target values for each project are set after the baseline data are checked and before the main validation, and favorable indicators are not picked after the results are seen.
What happens if the expected effect does not appear?
When scholarly review takes place, the meaning of failure also changes. When a system does not produce the expected effect, it can be examined separately whether this was an error of implementation, a mismatch in the conditions of application, or a limit of the original hypothesis. What the Scholars’ Council for AI seeks to build up is not only good results but a record of the evidence and judgments that make such distinctions possible.
How are requests from the organizations that fund research handled?
The roles of those who commission the research, those who carry it out and those who review it as scholars need to be made clear, and interests that could affect the results need to be recorded. Even when research results differ from expectations, it should be possible to review them. When the requests of the funding organization differ from the needs of the people actually affected, that difference is recorded and reviewed.
Can the results of one study be used as they are elsewhere?
When one applied study ends, the next question is how far its results generalize. AI research can also return questions to existing scholarship, but the laws of human society cannot be settled from results obtained in artificial systems alone. The outcomes of one site are not extended straight to effects on society as a whole; they are carried forward as questions to be checked again at the next site.
What is to be decided at the launch
- Deciding whose difficulties, and which ones, to help resolve
- Turning that problem into questions that can be verified
- Gathering the scholarship and data needed
- Setting participants’ contributions and the criteria for evaluation
After that, whether harm has decreased, whether people can understand and correct decisions, whether capability has grown, and whether a wider range of people can use the help are examined within the limits of the evidence.