Emerging Insights

Our Approach

Emerging Insights has developed a solution which allows companies to create combinable Knowledge and predictive, understandable, automatically updated models using Emergent law-based statistics.

This approach uses two fundamental empirical findings:

• We find lots of recurring patterns in data. These patterns can be formalized to what we call emergent laws.

• We empirically show that recurring patterns will keep repeating at a minimum rate again in the future. This can be used to create laws about the lower bound of emergent law-based prediction performance (meta laws).

These fundamental findings enable Emerging Insights to develop AI-solutions that are not viable with standard probability-based approaches. They allow clients not only to increase the performance of their predictive models but also to generate in-depth knowledge about the reasoning behind the model and factors influencing the results and to update this knowledge by objective falsification and updating rules.

Solutions

Emerging Insights developed a suite of emergent law based ML-algorithms to build models for prediction, decision and causal analysis.

_{Finding an emergent Law: In every sequence of at least 1500 car sales Volkswagen cars were on average more expensive than Renaults}

We are using the following advantages of the emergent law based methods:

• Patterns that were always empirically true are objective (no modelling assumptions are used).
• The modelling process can be guided by meta-laws.
• Always true empirical pattern cannot contradict each other.
• There exists an empirically justified decision rule: Choose the strategy that always leads to a better overall result after T decisions (T-Dominance).

They result in the following properties of our approach:

• Our methods are significantly easier to use correctly because we automate most model-building decisions through the use of meta-laws. These decisions usually required domain knowledge and a significant investment of time and resources by the data science team.
• The resulting models with up to date predictive performance are perfectly understandable and predictions are explainable.
• The T-Dominance decision rule leads to decisions that outperform conventional stochastic decision rules and allows us to build consistent multi-goal decision algorithms.
• Our software can create distributed databases collecting consistent empirical knowledge about dependencies between potentially millions of variables of interest (KnowledgeBases).
• These KnowledgeBases can then be used for causal analysis of business problems, e.g. quality assurance for production-systems or optimization of business processes.

Mission

The ability to find patterns that have always been true , to store, combine and update relevant empirical commonsenseworld knowledge in KnowlegeBases and to construct decision rules that always lead always to improved results leads enables us to build potentially very powerful tools. We are sure that the application of our algorithms and our methodology has the potential to change business processes, science and social decision-making substantially. So it is our mission is to spread the knowledge and use of our methods in a responsible way that leads to better informed decisions of institutions and people and to a sustainable improvement of human welfare.

Our Team

The founders of emerging insights spend more than 30 years in empirical research and the development of predictive models in university, industry and government consulting. The last ten years we spend most of our time in developing the methods and algorithm that are now ready to use for companies and institutions. The founders of emerging insights are: