Linus Behrbohm
@mankinskin

About me

Hi, my name is Linus and I am a Student of "cognitive informatics" at Bielefeld University.

My project

I am working on a web platform for collaborative work on projects of any domain. The goal is to have an open knowledge base and project planning tool for modelling problems and solutions.

The app is supposed to focus on an individual use case and should then be augmented by collaboration with others. The fundamental idea of the application is that individuals can create their own model of the real world as they see it and define their individual problems and goals as states of this model. Multiple people can compare their models and goals to learn from each other and collaborate for their shared goals in organized projects.

The heart of this idea is the formalization of states and relationships into a language which is capable of representing the real world as completely as possible, but still allows abstract modelling of every-day problems and solutions.

Goals are defined as assertions or tests on world states. These assertions are expressed in the formal modeling language and evaluate to a truth value, representing whether the goal is satisfied or not. You could for example define the goal that you must always have at least an average 75% score on your university assignments or that you must never spend more than you are earning.

Problems are defined by unsatisfied goals. When a goal is not satisfied, there are some values in the current world model which do not meet the goal's criteria. A solution to a problem is a series of changes to the world which end up in a state that matches a goal state, i.e. the goal is satisfied.

To model solutions we need to model time, because solutions are changes to the world which satisfy goals. Some changes to the world can be predicted fairly accurately, others can not. For example it is very likely that the sun will rise the next day and if we accurately model the solar system we can very accurately predict this event and the continuous change that will lead up to it. However, because our model will presumably never have complete, perfect information, there are a lot of events we can not accurately predict at all. We can always try to make predictions based on recorded events, but many relationships are simply impossible for us to infer, because not all state changes can be recorded.

One important example of difficult to predict events are human actions. Humans, or even other higher animals are such complex systems, that it will be incredibly difficult to predict their actions for a long time to come. But our world model still needs to be able to formalize the possible actions a human can take, without the requirement of an explicit explanation. The system needs to be tolerant to incomplete models of world events and should be able to make assumptions about the missing information to retain consistency with the model. The only thing which is not allowed is to introduce entirely contradicting information.