The title of a new inquiry from the US Federal Communications Commission (FCC) speaks for itself.
Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts
It does not require much imagination to think of ways that AI could affect the communications we receive. Robocalls and robotexts have already forced many people to change how they behave. The ringing of a telephone used to generate curiosity about the identity of the person who called us; now it can prompt dread of yet another machine-instigated sales pitch. The FCC often observes that unwanted robocalls are the leading cause of consumer complaints. They are less keen to examine why this has remained the situation for many years, despite the expensive and flawed solutions the FCC has foisted upon the communications industry. These typically get described as a ‘step in the right direction’ although statistics indicate that the number of unwanted robocalls received by Americans is rising. Artificial intelligence could turbo-charge the machinery available to scammers, or it could provide the technological miracle cure that the FCC keeps seeking because prosecutors do not secure meaningful punishments for the criminals behind illegal calls. Will AI be a savior or a curse?
The potential of AI is so great and our ability to predict the pace of technological progress is so limited that the FCC’s consultation document reads like a long laundry list of questions. It is so long that attempting to summarize their questions would be counterproductive; any question excluded from the summary may subsequently prove to be the most important of all. I cannot tell you how AI will evolve, so I will instead focus on a subject that can be more effectively analyzed using historical data: the strengths and weaknesses of regulators. Here are my three key takeaways from how the FCC chose to frame its conversation about AI and unwanted machine communications.
The FCC Should Waste Less Time Trying to Define AI
Lawyers tend to get fixated on laws, and laws are ultimately words. The lawyers who run the FCC dedicate eight paragraphs of their consultation document to comparisons of how different parties define AI. This might be appropriate if anybody expects the law to be used to punish the misuse of AI. However, the FCC and the US legal system have demonstrated over many years that they are incapable of using the law to punish the misuse of far simpler robocalling technologies. A legally watertight definition of AI is unnecessary if you are asking people to suggest practical methods of using AI or countering the uses of AI. Practical people are concerned with goals and how to achieve them. Goals like ‘identify patterns indicative of unwanted calls’ or ‘protect people who think they are talking to a person when they are actually talking to a machine’ can be usefully stated without the precision needed if they were going to be challenged in a court of law.
The Benefits of AI… But for Whom?
It is no surprise how the FCC has framed press releases about the way AI might influence the topic that causes most consumer complaints. This is reinforced by the first sentence in the section of their consultation document that asks about the potential benefits of AI.
We seek comment on how AI technologies may be used to protect consumers from unwanted and illegal robocalls and robotexts.
But then the FCC shows how it favors a section of the US business community by immediately pivoting to a question which is quite different.
Specifically, we seek comment on how AI technologies can be used to target information to specific groups of consumers…
This sounds a lot like the FCC trying to rationalize how businesses might reduce the total number of robocalls by discovering ways to improve the targeting of future robocalls. Is there any evidence that consumers want better targeting of unsolicited calls and messages, as opposed to them simply demanding fewer unsolicited calls and messages? Improved targeting is the holy grail of marketing. A follow-on question continues the theme of linking consumer protection to enhanced marketing efficiency.
Can AI technologies also benefit the calling party by enabling them to target demographics that are most interested in the subject matter of the communication?
The FCC asks many sensible questions about how AI could be used to reduce the harm and annoyance caused by unwanted and illegal calls and messages. However, the single factor which is most responsible for the FCC’s past mistakes is the excessive influence granted to organizations who want to make unsolicited sales calls. That same fault is evident here too.
The Big Risk That Gets Only One Sentence
The consultation document outlines many risks that are worthy of consideration, such as the potential to use AI to deceitfully impersonate human beings or the use of AI to vary traffic patterns in order to defeat anti-spam controls. However, there is only one sentence about the following risk.
Is there a risk that AI technologies might be used in such a way as to inadvertently block legitimate calls and messages?
As written, this question is facile. It should already be obvious that this risk exists; there is not a single competent technologist who denies the possibility that AI will make mistakes. If there were no risks than we might well ask why this foolproof AI is not already being used to solve the much simpler problem of correctly filtering email. But what makes this lone sentence worrying is the lack of any exploration of the well-known concern that AI decisions may be systematically biased against some segments of the population. The FCC’s reluctance to discuss the possibility that AI could block calls and messages in a way that is biased by the phone user’s race, gender, age or language contrasts sharply with the content of another document which the FCC references at the beginning of their consultation paper: President Biden’s ‘Blueprint for an AI Bill of Rights’. Biden’s AI blueprint is much shorter than the FCC’s AI consultation document, but note how often the White House discusses the risk of bias in the context of AI:
In America and around the world, systems supposed to help with patient care have proven unsafe, ineffective, or biased. Algorithms used in hiring and credit decisions have been found to reflect and reproduce existing unwanted inequities or embed new harmful bias and discrimination…
On his first day in office, the President ordered the full Federal government to work to root out inequity, embed fairness in decision-making processes, and affirmatively advance civil rights, equal opportunity, and racial justice in America…
Automated systems should be developed with consultation from diverse communities, stakeholders, and domain experts to identify concerns, risks, and potential impacts of the system… Outcomes of these protective measures should include the possibility of not deploying the system or removing a system from use…
You should not face discrimination by algorithms and systems should be used and designed in an equitable way. Algorithmic discrimination occurs when automated systems contribute to unjustified different treatment or impacts disfavoring people based on their race, color, ethnicity, sex (including pregnancy, childbirth, and related medical conditions, gender identity, intersex status, and sexual orientation), religion, age, national origin, disability, veteran status, genetic information, or any other classification protected by law…
Designers, developers, and deployers of automated systems should take proactive and continuous measures to protect individuals and communities from algorithmic discrimination and to use and design systems in an equitable way. This protection should include proactive equity assessments as part of the system design, use of representative data and protection against proxies for demographic features…
This framework describes protections that should be applied with respect to all automated systems that have the potential to meaningfully impact individuals’ or communities’ exercise of… Equal opportunities, including equitable access to education, housing, credit, employment, and other programs…
Note how much the White House has to say about the risk of bias and AI. Then think again about the single sentence in the FCC’s consultation which asks a banal question about whether AI ‘might’ be used to inadvertently block legitimate calls and messages. Blocking calls and messages can potentially impact many aspects of a person’s life, including their access to healthcare, education and jobs. It could even stop them talking to a loved one during an emergency. However, one of the most serious dangers with the FCC’s hope of using AI to block unwanted comms is barely acknowledged by their consultation paper. And the FCC has no excuse, because multiple parties have already warned them that existing anti-spam controls in the USA do a terrible job of distinguishing between legitimate and illegitimate traffic.
The full text of the FCC’s consultation document on AI and robocalls can be found here.
