[The following is Part II in a two-part guest post by Will Bridewell and Alistair M. C. Isaac. — JS]
Part 1 approached the problem of deception from a computational perspective, arguing that, in order to reason effectively about deception, an agent must be able to represent not only the beliefs and desires of a speaker but also the hierarchy of standing norms that govern her actions and speech. Deception occurs when an ulterior motive supersedes standing norms. This perspective has the advantage of correctly identifying forms of deceptive speech where the intent is not to cause a false belief, or where the utterance itself is not false, for instance, paltering. We also believe this approach has some novel implications for political discourse.
Much public political debate appears counterproductive — participants from either side of the aisle appear to “talk past” each other or to view rebuttals from their political opponents as non sequiturs. Crucially, these debates not only fail to progress but also fail to fulfill our intuitive notion of communication — participants’ original views become entrenched, rather than changing and developing in response to their interlocutors. Each party to the discussion interprets, or at least reacts to, the utterances of the other as inappropriate. Our basic framework suggests a diagnosis of this puzzling pattern of exchange.
On our view, appropriate speech is the result of constraint satisfaction over ranked standing norms, only one of which is be truthful. An implication of this view is that seemingly inappropriate speech may result from either a disagreement about matters of fact or a disagreement about norms and their ranking. Some counterproductive political debates appear to result from a discrepancy between the levels of discourse on both sides. An argument seemingly at the level of facts may mask an underlying norm conflict in one or both participants. For example, consider the following family argument between two siblings at the dinner table.
Andre: It’s just like the liberal media to post biased pictures of the crowd at the president’s inauguration.
Iman: Oh, come on. If you look at the numbers, you can see that there were fewer people than usual. Numbers don’t lie.
Andre: Those are your facts, not mine.
As voices amplify and tensions escalate, even the power of Wikipedia and online statistics cannot resolve what seems to be a simple question of fact (i.e., the number of attendees at various recent inaugurations). Hope for consensus and reconciliation seems lost.
This apparent argument about facts, their validity, and their relevance hits a barrier and seems to provoke irrationality. However, several norms may be in play, including epistemological ones such as preserve entrenched beliefs or maintain overall coherence, as well as political and social norms associated with in-group identification. Andre’s steadfast defense of a large attendance could be tied to norms that constrain not only belief preservation but also right political action. As a result, Iman’s appeal to facts, which may be guided by her desire to emphasize an official’s weakness, runs headlong into those political norms that structure Andre’s beliefs. If reorganizing or discarding normative constraints is more disruptive to one’s beliefs than rejecting an apparent fact, then both Andre and Iman are acting rationally. The irrational position is thinking that a single fact should have the power to disrupt a person’s hierarchy of norms.
A second example illustrates the converse scenario: a question of fact is obscured behind a debate about norms. Recently, several professional athletes have chosen to kneel during the national anthem as it plays before sporting events as a means to draw attention to a racial imbalance in police use of force. The key issues here are factual: “How frequently do police use force toward minority groups,” and, “Is there evidence that their behavior reveals racial bias?” Nevertheless, much of the public response to kneeling has focused on the norms of patriotism and appropriate behavior toward flags, anthems, and other national symbols. The accusations that these athletes are unpatriotic are meant to undermine the act of kneeling in protest.
This case reveals a paradox of political discourse. On the one hand, disenfranchised groups (by definition) have trouble influencing political debate and thereby altering societal norms. Flagrant norm violations may draw attention to the group’s plight within the public sphere — traditionally, this is the role of civil disobedience. On the other hand, the very use of norm violation as a means of initiating communication provides the empowered with a ready response: defend the violated norm, while ignoring both the facts that spurred its violation and the call to change the originally contested norm. Because the protestors’ appeals to facts target the norms of police behavior, they have no force against a response that only addresses norms for observing the national anthem.
What lessons can we draw from these examples? First, debates can occur at the level of facts, norms, or the ranking of norms — patriots may kneel during the national anthem, if they take some other value (e.g., individual liberty) to supersede patriotism in that context. As a result, if debate is to proceed constructively, the parties involved may need to explicitly ask whether they (a) want to have shared values, (b) can identify differences in preferences, and (c) can determine where the source of conflict resides. This discursive strategy is common in negotiations, where the actual desires (as opposed to the stated desires) of the groups involved are often unknown until brought into conflict. Second, without some, perhaps mediated, process of discovery, political “debate” may merely re-enforce existing power structures. In the dinner table conversation, for instance, vacuous debate over an inconsequent question of fact merely deepens commitment to the underlying, undiscussed norms.
What then is the role of ulterior motives and norms in the political sphere? Without the ability to represent norms and the preferences among them, there is little hope of identifying the actual locus of disagreement in heated discussions like those above. That is, the act of confusing arguments about facts with arguments about values may lead to a stalemate or, even worse, unintentional sabotage of one’s own position. Regardless of whether we find ourselves in a deceptive scenario, a political argument, or a heated negotiation, we have to know not only what the facts are, but also how what is being said connects with the norms, goals, and beliefs of the people involved. A representational framework like the one that we have developed for detecting deception may aid this process by providing a means to make these connections explicit, as they must be for debate to proceed rationally.
Will Bridewell is a computer scientist at the U.S. Naval Research Laboratory’s Navy Center for Research in Artificial Intelligence. Alistair M. C. Isaac is a lecturer in the philosophy of mind and cognition at the University of Edinburgh. Their computational approach to deception is explored in more detail in “White Lies on Silver Tongues: Why Robots Need to Deceive (and How)” in Robot Ethics 2.0, Oxford University Press.