Artificial intelligence (AI) has become the latest technology shaking up and reshaping the unified communications (UC) space, and with it comes new questions and concerns. How can companies responsibly and ethically use AI in their platforms? How can we program AI to ensure privacy? Can AI be biased? All these questions and more are continuing to drive conversations and must be addressed by any businesses looking to use AI.
With that in mind, here are some of the top ethical considerations around AI in unified communications, the problems that arise, and how companies can address them.
Transparency: How is AI Being Used?
One of the first key questions is the issue of transparency. AI can be a useful tool for providing assistance during calls, taking notes and transcripts, and pulling up information at a moment’s notice, but even when AI feels like a seamless part of the communication experience, it still needs to be clear when it’s being used.
While many AI developers boast about how talking with their AI feels like talking with a human being, users still need to know when they’re talking with a bot and when they’re talking with a person. If an AI bot is being used for something like customer self-service, then it’s even more important that the customer understands they’re not speaking with a real person.
Users need to understand when generative AI is being used, how it works, and what it’s used for. For instance, companies can include a sidebar explaining what their AI does, where its training data comes from, and how it can be used. In text chats, AI responses and suggestions can appear in a new, clearly labeled tab, or as an AI participant—as long as it’s clear that it’s a bot.
The important thing is that companies are transparent about their use of generative AI and clearly explain its uses.
Privacy: What Does the AI Know?
One major concern about AI is where its data comes from. Since AI can only use what data it’s fed, it has to rely on massive amounts of training data. In many cases, this can include existing articles, websites, and other publicly available sources, as well as internal training data for specific companies.
However, AI also tends to learn as it is used, which means it’s constantly collecting new data. Companies need to make sure that their AI is not collecting or using personal data in ways that may violate privacy laws or otherwise be intrusive—especially if the AI is used in industries like healthcare or finance, where important personal information is shared and protected under strict regulations.
There needs to be a clear separation between conversations and data an AI can learn from, and that which it needs to delete to avoid compliance issues. If the AI learns typical company jargon, that’s one thing, but if it starts learning users’ identifying information, that can be a problem, especially if it shares it with other users (intentionally or otherwise). Not only does personal information need to be protected, but employees need to feel confident that the AI is only used for enhancing communication, not for monitoring them.
Bias and Discrimination: Is the Data Diverse Enough?
Bias in AI is a major issue, which companies need to be aware of every time they provide training data. Generative AI is only as good as the data it’s trained on, which means that if the data is biased, the AI will be too. This can often have huge unintended consequences, ranging from social or systemic biases that discriminate against groups, to data sampling that isn’t properly representative of all groups, to implicit biases skewing data.
For instance, Amazon recently made headlines for its AI recruiting tool that showed a bias against women. Because the AI was trained to identify patterns in resumes, it identified that a majority of tech resumes came from men—which then led it to erroneously assume that men were better qualified for tech jobs than women.
Errors like this are often caused by a lack of diverse and properly representative data. Companies need to make sure that the data they give the AI properly represents diverse groups equally, and does not draw erroneous conclusions based on race, gender, or any other identities.
Generative AI must not be used to discriminate against any individuals or groups. This requires careful monitoring of the data it’s fed and the conclusions it draws from them to avoid any biases that may result in discriminatory practices or behavior. Even unconscious biases can sometimes make their way into the data sets that are used for AI training, so companies need to be aware and stay on the lookout to avoid a situation like the one caused by Amazon’s recruitment AI.
