I am excited to share the introduction of a new SAP HANA Cloud Machine Learning feature: Fairness in Machine Learning (FAIR ML). FAIR ML will be added to the
Predictive Analysis Library (PAL) in Q4 of 2023 and will help mitigate bias and unfairness in ML models that result from data related to features such as gender, race, age, or other protected classes.
FAIR ML will be offered with hybrid gradient boosting tree (HGBT) binary classification and regression models.
Acknowledging the Societal Impact of Technological Advancement
Aiming for fairness and ethical business practices is a crucial objective for society. With technology learning from data derived from humans, a new challenge now becomes the bias being passed, reinforced, and potentially even amplified by technology such as artificial intelligence (AI) and ML.
As outlined in
SAP’s Global AI Ethics Policy, some of the key activities resulting in AI and ML systems contributing to bias and discrimination are by:
- Gaining insights from the existing structure and behaviour of the societies they analyze and thereby reproducing, reinforcing, and amplifying patterns of marginalization, inequality, and discrimination that exist in society and may be encoded into data sources used for the creation of AI.
- Replicating the developers’ preconceptions and biases since many of the features, metrics, and analytic structures of the models that enable data mining are chosen by developers.
- Training and testing algorithmic systems on data samples that may be insufficiently representative of the populations or the past situations from which they are drawing inferences.
Former president of SAP SuccessFactors, Jill Popelka, established that technology has a leading role to play in helping organizations live and work by diversity, equity, and inclusive practices in this
press release.
SAP’s Guiding Principles for Artificial Intelligence present a framework to steer AI software to
help the world run better and improve people’s lives. These guidelines represent a commitment to move
business beyond bias and what is legally required in order to facilitate a deep and continuous engagement with the wider ethical and socioeconomic challenges of AI.
Innovating with Fair Machine Learning Models
SAP HANA Cloud enables customers to unleash the full potential of data through consolidation for multi-model and ML processing in a single database. The embedded ML allows customers to leverage their private business data to optimize organizational operations and build intelligent data applications. With the addition of FAIR ML, data scientist and application developers can use SAP HANA Cloud to build ML models that mitigate unfairness with respect to human sensitive data. This algorithm will decrease disparities, avoid harm, and assure fairness in decision-making. Leveraging the FAIR ML function may avoid bias to any particular group with AI-augmented decisions on humans. Example use cases where the FAIR ML function can play a pivotal role include college admission, job candidate selection, or personal credit evaluation.
Unlock Machine Learning in SAP HANA Cloud
To learn more about the FAIR ML function, check out Xin Chen’s
blog where she goes through a case study highlighting fairness-related harms introduced in a ML model, and how they are mitigated with the new feature. Additionally, to gain an understanding of how ML works in the context of building intelligent data applications, check out Susan Poppe’s
blog.
To experience ML within SAP HANA Cloud, check out
this blog highlighting resources such as the guided experience. For specific questions, schedule a talk with a
product expert or visit the
SAP HANA Cloud community where users and experts collaborate on product questions regularly.