Salesforce takes a principled approach to ethical AI to maintain trust

Do you get followed around by the same ad?

Have you experienced an ad that creepily follows you around? The ad follows you from the desktop to your mobile or iPad, trying to entice you to buy a product you have no interest in. Creepy isn’t it?

You might have inadvertently stumbled across their website or commerce platform, and now, like chewing gum stuck to your shoe, it’s impossible to shake it off. It can even feel somewhat sinister as if some sneaky piece of code or spyware has been dropped on your device.

The seller might be shocked that you could suspect it of being underhand or unethical.  For decades marketers have tried to deliver more personalized campaigns to boost conversions. Initially basing their efforts on crude segmentation techniques and assumptions about the demographic that is most likely and able to buy.

Things have, of course, moved on from these former crude attempts. Marketers now talk about personalization and more granular segmentation aided by better data and, more recently, AI in the form of machine learning and sentiment analysis.  Before AI became mainstream, there were concerns that perhaps in the distant future, AI would advance so far that eventually, it could dispense with data scientists, creating evermore advanced versions of itself that no one can control. This fear seems to have subsided in the last few years, although regulators keep a watchful eye on AI developments.

Ethical AI is the immediate challenge

There is, however, a more concerning development with AI, and that is the potential for bias baked into the underlying algorithms. Regulators like the US Federal Trade Commission (FTC) and the European Commission (EC) continue to grapple with this, and both recognize the potential upside of AI to boost commerce, health, and environmental sustainability. They also recognize the potential harm, however innocently intended, for AI to impact decisions based on embedded algorithmic bias.

The FTC actively monitors AI developments and their potential for unethical use:

  • Section 5 of the FTC Act. The FTC Act prohibits unfair or deceptive practices. That would include the sale or use of – for example – racially biased algorithms.
  • Fair Credit Reporting Act. The FCRA comes into play in certain circumstances where an algorithm is used to deny people employment, housing, credit, insurance, or other benefits.
  • Equal Credit Opportunity Act. The ECOA makes it illegal for a company to use a biased algorithm that results in credit discrimination on the basis of race, color, religion, national origin, sex, marital status, age, or because a person receives public assistance.

Beyond the GDPR regulations, the EC has proposals in place to strike a balance between the potential benefits of AI and the protection of human rights. The proposed regulatory framework on Artificial Intelligence has the following  objectives:

  • Ensure that AI systems placed on the Union market and are safe and respect existing laws on fundamental rights and Union values
  • Ensure legal certainty to facilitate investment and innovation in AI
  • Enhance governance and effective enforcement of existing law on fundamental rights and safety requirements applicable to AI systems
  • Facilitate the development of a single market for lawful, safe, and trustworthy AI applications and prevent market fragmentation.

Salesforce’s approach to ethical AI builds on its core values

I last looked at the subject of AI from an ethical standpoint back in February this year.

And given the ubiquity of the Salesforce Customer 360 CRM platform, it was with great interest that I attended an analyst briefing on Salesforce’s approach to ethical AI.

Salesforce’s core values are the foundation for its approach, the most crucial being – trust. In addition, five guiding principles govern the development and ethical use of AI in its products. The five principles are:

  • Human rights – to ensure the direct use of the vendor’s technologies uphold equal and inalienable protections.
  • Privacy – providing best practice approaches to product design, enabling enterprises to protect each customer’s data.
  • Safety – protecting people from direct harm from the use of the technology.
  • Honesty – opposing the use of its technology from spreading disinformation or conspiracy theories and providing tools to ensure AI is used responsibly.
  • Inclusion – by providing equal access to technology.

The diagram at the top of this post outlines a simple ethical AI practice maturity model. Many organizations have gone from an initial ad hoc approach to ethical AI with few policies in place to govern AI to the most advanced systemic, optimized and innovative use of AI. In the most progressive state, products are blocked from release if ethical risks surface from key metrics, such as indications of bias.

Another Salesforce core value is innovation. Salesforce follows an Agile development methodology that addresses ethical risks from start to launch and post-launch monitoring. By applying an ethical lens throughout the development process, potential risks are identified and mitigated.

Rather than slowing up product development, this disciplined development approach improves code quality based on the principle of a stitch in time saves nine.

Built-in safeguards prevent AI abuse

Salesforce has embedded several safeguards in Einstein to prevent unethical AI. This includes:

  • Changeable gender-based spectrum field – gender-inclusive attributes such as non-binary, prefer not to say in its Nonprofit Success Pack
  • Einstein content selection  – which scans names and attribute values for sensitive features that might introduce bias.
  • Explainability features (transparent AI) to describe why a decision was made and the factors that influenced it.
  • Removal of cultural bias correlated with cultural attributes within the Bot model.
  • Model cards to describe model inputs, outputs, and scope of use cases.

Ultimately each organization must develop its own policies and risk mitigation plans

It’s reassuring to know that Salesforce takes ethical AI very seriously. However, ultimately it is the responsibility of the leadership within each company to put in place the policies and best practices to deliver a trusted environment for customers. Where on the ethical AI maturity model does your company lie? If not at the optimized and innovative end of the spectrum, formulate a plan to embed ethical practices throughout your organization. The trust of your customers is at stake.

Previous

Next