Algorithmic bias

Technology can amplify existing biases. "One of the more pressing ethical issues facing today's developers is bias," said Spencer Lentz, principal, AI and digital process automation, digital customer experience, at consulting firm Capgemini.

Bias often enters the system undetected -- Lentz compares bias to a virus. Computers themselves have no inherent moral framework. Software can only learn from the training it is given. Therefore, developers and data scientists must scrub bias from the training data and the algorithms they build. From a developer's perspective, bias often centers on eliminating options for the wrong reasons, Lentz said.

Reporting and research in recent years illustrates how bias within software systems can perpetuate systemic racism against specific populations, which creates lost opportunity, worsens medical care and increases rates of incarceration. For example, in the book, Race After Technology, Ruha Benjamin raised some concerns about a case where developers failed to include Black people's voices in training AI speech recognition algorithms, under the belief that fewer Black people would use the app.

Anaconda, a data science platform, conducts the "State of Data Science Survey" and in 2020 found that 27% of data practitioners thought the biggest problem to tackle in AI and ML was the social impact from bias in data and models.

"To reduce bias in data and models, practitioners must be intentional about their work, asking questions like 'How was this data collected,' and 'What assumptions were collected with it,' said Peter Wang, CEO and co-founder of Anaconda. "If you don't know how the data that trained your models was collected and under what conditions, then your models could inadvertently perpetuate biases in their outputs." Executives, data scientists and developers must create an organizational culture that establishes ethical guidelines and empowers individuals at any level of the business to speak up if they see something problematic.

It's time to create a governing body, similar to the American Medical Association for doctors, Wang argued. This body could establish industry-wide ethical guidelines and best practices. "These technologies are still relatively new in the business context, and we would all benefit from ethical standards derived from our collective intelligence and input, rather than leaving it up to each individual or organization to decide for themselves," he said.

Question: 

In less than 400 words, identify and apply the PAPA framework for ethical issues which appear in the following text.

Answer rating: 100% (QA)

PAPA framework The framework is designed to deal with the privacy accuracy property and availability ... View the full answer