The case for responsible personalization

The current state of online privacy presents a lot of uncertainty for both consumers and businesses. The recent explosion in AI innovation has brought a new range of security and privacy threats, from deep fakes, voice clones and the like, to concerns such as how sensitive business information is handled when working with generative AI.

This is one major reason why, on the one hand, people have become much more aware of how their data is used and distributed, making them more cautious with how they share their data and how their online privacy can be violated.

On the other hand, however, people have become so used to personalized experiences that they’re now a prerequisite rather than a nice to have, and their online data is still the most reliable method for effective personalization, with AI technologies playing a key role as well.

So, how can marketers balance these two seemingly contradictory market trends while also taking into account the new privacy challenges posed by AI?

Enter responsible personalization, which takes into account the privacy concerns of individuals and businesses while making sure that data and AI are handled responsibly.

Responsible, privacy-first personalization

The general public has recently become much more conscious of the importance of online privacy. Some key factors here were the privacy scandals that led to stricter regulations such as GDPR, and the fact that the topic of privacy was tackled in popular culture by Netflix in their documentary film “The Social Dilemma”.

Now it is widely recognized that third-party cookies, mainly used for advertising purposes, are the major source of privacy concerns. While Google has recently made the decision not to deprecate third-party cookies after all, that doesn’t mean you need to keep relying solely on them to deliver great personalized experiences. Let’s take a look at how you can use first-party data for personalization.

First-party data is data obtained from your own channels rather than from third parties, i.e. from your social media profiles, surveys and forms on your website(s) and apps, etc. Basically, first-party data is data your visitors and customers share freely with you in exchange for a better online experience with your brand.

Another approach to consider is using manual over automated personalization. In B2B, this may look something like typing out a sales email (without a template) to a prospective client you’ve met in person at a conference. In B2C, customer service often includes chatbots and human customer service representatives working in tandem to deliver the best experience.

While it may seem like third-party data would be inherently more valuable in B2C than first-party data, the latter may actually provide more value and options than it might initially seem.

First-party data is often obtained in exchange for exclusive perks such as discounts or special items (or special content such as an extensive whitepaper in B2B). This is a big win-win since you’re able to learn more about your customers while also giving them perks that are tailored to the information you already have about them, either from previous first-party sources or from third parties.

To return to third-party data – since third-party cookies are not getting deprecated on Google Chrome, they will most likely remain a mainstay of the web and hence a mainstay of web personalization.

To ensure as much privacy as possible even with third-party cookies, you need to be transparent about how you collect data and for what purposes you use the collected data. Make sure to follow GDPR (and/or any other privacy regulation which applies to your business) and the proposed standards for the cookie & privacy notices as well as the privacy disclaimer on your website (remember: it should be opt in, and not opt out).

Responsible, ethical AI

How you collect and use data from your customers is also essential for responsibly working with artificial intelligence. In fact, any AI implementation and/or integration should follow the standards of ethical AI. This means accounting for all potential risks, both short term and long term, which come with a particular AI implementation, and always considering the context in which AI is used.

One of the main issues with unethical AI is bias propagation, typically based on the underlying data used to feed the AI system which is itself often biased. The second issue is inherent to how generative AI actually works.

Since these systems generate language by scraping existing web content and then making predictions based on this existing content in order to generate new content, this new content can often seem – or actually be – plagiarized. And, since specific versions of large language models include web content up to a certain date, this means LLMs often work with outdated information, leading to AI hallucinations.

But even a well-intentioned, responsibly implemented AI system can be misused or abused; in extreme cases, this can be done by bad actors that intentionally use AI for unethical, nefarious purposes.

Luckily, less severe cases of AI misuse can be more or less avoided with the proper approach to AI, beginning with the right training to teach employees how to use AI more responsibly.

One of the main examples that we’ve seen of unintentional AI misuse is compromising sensitive business information by entering it into a publicly available LLM such as ChatGPT.

Granted, a mistake like this is very often the best failsafe against similar issues happening in the future, but preemptive training should still be the priority.

Companies and organizations should emphasize this and help employees become more careful in what kind of data they enter into publicly available tools. Conversely, they can also create private internal integrations of a publicly available LLM which would allow employees to work with sensitive business data more freely and securely.

Final thoughts

Given the rising concerns surrounding online privacy, it’s high time for businesses and organizations to adopt more responsible approaches to personalization. And, with AI becoming such a ubiquitous part of online experiences and often being key to automated personalization, responsible AI use is also a must for effective responsible personalization.