Managing Bias in Machine Learning Projects

Fahse, Tobias; Huber, Viktoria; Giffen, Benjamin van

doi:10.1007/978-3-030-86797-3_7

Cited by 16 publications

(9 citation statements)

References 38 publications

(73 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There were several papers that extend CRISP-DM in different dimensions. The studies in Kolyshkina & Simoff (2019) and Fahse, Huber & van Giffen (2021) addressed two important aspects of ML solutions—interpretability and bias, respectively. They suggested new activities and methods integrated in CRISP-DM steps for satisfying desired interpretability level and for bias prevention and mitigation.…”

Section: Resultsmentioning

confidence: 99%

Current approaches for executing big data science projects—a systematic literature review

Saltz

Krasteva

2022

PeerJ Computer Science

View full text Add to dashboard Cite

There is an increasing number of big data science projects aiming to create value for organizations by improving decision making, streamlining costs or enhancing business processes. However, many of these projects fail to deliver the expected value. It has been observed that a key reason many data science projects don’t succeed is not technical in nature, but rather, the process aspect of the project. The lack of established and mature methodologies for executing data science projects has been frequently noted as a reason for these project failures. To help move the field forward, this study presents a systematic review of research focused on the adoption of big data science process frameworks. The goal of the review was to identify (1) the key themes, with respect to current research on how teams execute data science projects, (2) the most common approaches regarding how data science projects are organized, managed and coordinated, (3) the activities involved in a data science projects life cycle, and (4) the implications for future research in this field. In short, the review identified 68 primary studies thematically classified in six categories. Two of the themes (workflow and agility) accounted for approximately 80% of the identified studies. The findings regarding workflow approaches consist mainly of adaptations to CRISP-DM (vs entirely new proposed methodologies). With respect to agile approaches, most of the studies only explored the conceptual benefits of using an agile approach in a data science project (vs actually evaluating an agile framework being used in a data science context). Hence, one finding from this research is that future research should explore how to best achieve the theorized benefits of agility. Another finding is the need to explore how to efficiently combine workflow and agile frameworks within a data science context to achieve a more comprehensive approach for project execution.

show abstract

Section: Resultsmentioning

confidence: 99%

Current approaches for executing big data science projects—a systematic literature review

Saltz

Krasteva

2022

PeerJ Computer Science

View full text Add to dashboard Cite

show abstract

“…For instance, [53] considers the machine learning lifecycle in seven steps and identifies seven distinct sources of bias. Fahse et al [54] identifies eight distinct types of bias and allocates them to the six phases of the CRISP-DM model. We use these inputs as we frame checklists for each layer of our model.…”

Section: Ai Lifecycle -Bias and Auditingmentioning

confidence: 99%

A seven-layer model with checklists for standardising fairness assessment throughout the AI lifecycle

Avinash

Harsh²

2023

AI Ethics

View full text Add to dashboard Cite

Problem statement: Standardisation of AI fairness rules and benchmarks is challenging because AI fairness and other ethical requirements depend on multiple factors, such as context, usage case, type of the AI system, and so on. In this paper, we elaborate that the AI system is prone to biases at every stage of its lifecycle, from inception to its use, and that all stages require due attention for mitigating AI bias. We need a standardised approach to handle AI fairness at every stage. Gap analysis: While AI fairness is a hot research topic, a holistic strategy for AI fairness is generally missing. The AI fairness assessment is often developer-centric. Standard processes are required for independent audits for the benefit of other stakeholders. Also, most researchers focus only on a few facets of AI model-building. Peer review shows a focus on biases in the datasets, fairness metrics, and algorithmic bias. In the process, other aspects affecting AI fairness get ignored. The solution proposed: We propose a comprehensive approach in the form of a novel seven-layer model, inspired by the Open System Interconnection (OSI) model, to standardise AI fairness handling. Despite the differences in the various aspects, most AI systems have similar model-building stages. The proposed model splits the AI system lifecycle into seven abstraction layers, each corresponding to a well-defined AI model-building or usage stage. We also provide checklists for each layer and deliberate on potential sources of bias in each layer and their mitigation methodologies. This work will facilitate layer-wise standardisation of AI fairness rules and benchmarking parameters.

show abstract

“…When this input is biased, this bias can be transferred into the model generated by the algorithm, leading to unfair decisions and a reduction in quality [25,26]. These problems can have severe financial, social, and reputational effects on companies [8].…”

Section: Bias In Machine Learningmentioning

confidence: 99%

“…There are various types of general biases happening in different stages of CRISP-DM, a well-known standard process for data mining introduced by IBM in 2015, which breaks the data process into six different stages including business understanding, data understanding, data preparation, modeling, evaluation, and development. The types of described biases are social bias, measurement bias, representation bias, label bias, algorithmic bias, evaluation bias, deployment bias, and feedback bias [8]. Social bias happens when the data transfer biases in society to the model on a large scale.…”

Section: Bias In Machine Learningmentioning

confidence: 99%

See 1 more Smart Citation

A Comparative Analysis of Bias Amplification in Graph Neural Network Approaches for Recommender Systems

2022

View full text Add to dashboard Cite

Recommender Systems (RSs) are used to provide users with personalized item recommendations and help them overcome the problem of information overload. Currently, recommendation methods based on deep learning are gaining ground over traditional methods such as matrix factorization due to their ability to represent the complex relationships between users and items and to incorporate additional information. The fact that these data have a graph structure and the greater capability of Graph Neural Networks (GNNs) to learn from these structures has led to their successful incorporation into recommender systems. However, the bias amplification issue needs to be investigated while using these algorithms. Bias results in unfair decisions, which can negatively affect the company’s reputation and financial status due to societal disappointment and environmental harm. In this paper, we aim to comprehensively study this problem through a literature review and an analysis of the behavior against biases of different GNN-based algorithms compared to state-of-the-art methods. We also intend to explore appropriate solutions to tackle this issue with the least possible impact on the model’s performance.

show abstract

Managing Bias in Machine Learning Projects

Cited by 16 publications

References 38 publications

Current approaches for executing big data science projects—a systematic literature review

Current approaches for executing big data science projects—a systematic literature review

A seven-layer model with checklists for standardising fairness assessment throughout the AI lifecycle

A Comparative Analysis of Bias Amplification in Graph Neural Network Approaches for Recommender Systems

Contact Info

Product

Resources

About