Data Quality
Measuring, managing and improving data quality in a value-driven way
In the wake of digitalisation, organisations today have access to vast and ever-growing volumes of data (keyword: ‘Big Data’). However, empirical evidence shows that the data analysed and utilised is often characterised by poor data quality – even in internal corporate customer databases, on average around 30% of the stored data values are incorrect. This results, for example, in additional annual costs of 15 million dollars for the average American company. However, poor data quality is not only a major problem for businesses – in politics and society too, the need for reliable information is growing in the age of ‘fake news’. Quantitative methods are therefore needed to measure, control and improve data quality.
This issue has particularly critical implications in the age of Artificial Intelligence (AI), where poor data quality leads directly to uncertainty. If these underlying uncertainties in AI forecasts are ignored – for example, by simply replacing missing values with averages – AI systems often issue overly confident and sometimes incorrect recommendations. For human decision-makers in human-AI teams, this poses enormous risks: A deceptive illusion of certainty arises, which can lead to overreliance and potentially serious misjudgements. To harness the transformative potential of AI in a responsible and trustworthy manner, it is therefore essential to quantitatively capture uncertainties arising from data quality defects, map them transparently and disclose them to users.
#DataQuality #FakeNews
Projects
- Data Quality in the Automotive Industry
- Data Quality in User-Generated Content (DQNGI)
- Data Quality in User-Generated Content (DQUGC)
- Data Quality Measurement and Measures for Wikis and Knowledge Graphs (DQMM@Wiki)
- Measuring and Improving Data Quality in Unstructured Data (DQMM)
- Making uncertainties visible: Uncertainty-aware AI