The purpose of this thesis has been to evaluate if a new instance based explanation method, called Automatic Instance Text Classification Explanator (AITCE), could provide researchers with insights about the predictions of automatic text classification and decision support about documents requiring human classification. Making it possible for researchers, that normally use manual classification, to cut time and money in their research, with the maintained quality. In the study, AITCE was implemented and applied to the predictions of a black box classifier. The evaluation was performed at two levels: at instance level, where a group of 3 senior researchers, that use human classification in their research, evaluated the results from AITCE from an expert view; and at model level, where a group of 24 non experts evaluated the characteristics of the classes. The evaluations indicate that AITCE produces insights about which words that most strongly affect the prediction. The research also suggests that the quality of an automatic text classification may increase through an interaction between the user and the classifier in situations with unsure predictions.

Explainable AI (XAI) has grown as an important field over the years. As more complicated AI systems are utilised in decision-making situations, the necessity for explanations for such systems is also increasing in order to ensure transparency and stakeholder trust. This study focuses on a specific type of explanation method, namely counterfactual explanations. Counterfactual explanations provide feedback that outlines what changes should be made to the input to reach a different outcome. This study expands on a previous dissertation in which a proof-of-concept tool was created for comparing several counterfactual explanation methods. This thesis investigates the properties of counterfactual explanation methods along with some appropriate metrics. The identified metrics are then used to evaluate and compare the desirable properties of the counterfactual approaches. The proof-of-concept tool is extended with a properties-metrics mapping module, and a user preference-based system is developed, allowing users to evaluate different counterfactual approaches depending on their preferences. This addition to the proof-of-concept tool is a critical step in providing field researchers with a standardised benchmarking tool.

<p> </p><p>The international maritime industry is growing fast due to an increasing number of transportations over sea. In pace with this development, the maritime surveillance capacity must be expanded as well, in order to be able to handle the increasing numbers of hazardous cargo transports, attacks, piracy etc. In order to detect such events, anomaly detection methods and techniques can be used. Moreover, since surveillance systems process huge amounts of sensor data, anomaly detection techniques can be used to filter out or highlight interesting objects or situations to an operator. Making decisions upon large amounts of sensor data can be a challenging and demanding activity for the operator, not only due to the quantity of the data, but factors such as time pressure, high stress and uncertain information further aggravate the task. Bayesian networks can be used in order to detect anomalies in data and have, in contrast to many other opaque machine learning techniques, some important advantages. One of these advantages is the fact that it is possible for a user to understand and interpret the model, due to its graphical nature.</p><p>This thesis aims to investigate how the output from a Bayesian network can be explained to a user by first reviewing and presenting which methods exist and second, by making experiments. The experiments aim to investigate if two explanation methods can be used in order to give an explanation to the inferences made by a Bayesian network in order to support the operator’s situation awareness and decision making process when deployed in an anomaly detection problem in the maritime domain.</p><p> </p>

The international maritime industry is growing fast due to an increasing number of transportations over sea. In pace with this development, the maritime surveillance capacity must be expanded as well, in order to be able to handle the increasing numbers of hazardous cargo transports, attacks, piracy etc. In order to detect such events, anomaly detection methods and techniques can be used. Moreover, since surveillance systems process huge amounts of sensor data, anomaly detection techniques can be used to filter out or highlight interesting objects or situations to an operator. Making decisions upon large amounts of sensor data can be a challenging and demanding activity for the operator, not only due to the quantity of the data, but factors such as time pressure, high stress and uncertain information further aggravate the task. Bayesian networks can be used in order to detect anomalies in data and have, in contrast to many other opaque machine learning techniques, some important advantages. One of these advantages is the fact that it is possible for a user to understand and interpret the model, due to its graphical nature. This thesis aims to investigate how the output from a Bayesian network can be explained to a user by first reviewing and presenting which methods exist and second, by making experiments. The experiments aim to investigate if two explanation methods can be used in order to give an explanation to the inferences made by a Bayesian network in order to support the operator’s situation awareness and decision making process when deployed in an anomaly detection problem in the maritime domain.