Deception Detection Techniques for Journalism

kentbye's picture
| | | | |

Johnson et. al. provides a taxonomy for deception tactics and suggests heuristics to counter each of the tactics. I've only briefly summarized or pointed to Johnson et al's deception detection technique. I thought that it'd be useful if I provided a rough map to some of the insights from their article, and how they could be used in a citizen journalistic context -- or any other domain with the potential for human deception. All references can be found in Johnson et al's article.

I'll just give an overview of the technique and you can go and read the details more in their paper -- the juicy stuff starts on page #359 with the section called "A theory of detecting deception in financial information."

They use the auditing domain of financial deception where the corporate management knowing falsifies financial statements in order to deceive the auditor. I have replaced "corporate management" with "political official" as the deceiver, and "auditor" with "journalist" as the deception target.

In order for a politician to create a "Deception Core" that deceives a journalist, the first step is to analyze how a journalist thinks.

Johnson et al. uses Neisser's ('76) Deceiver's Theory of Target's Mind (i.e. the Journalist's Mind):

1.) Journalist seeks info from environment (asks questions, observe political behavior)

2.) Journalists categorizes or interprets this info (mentally processes info and reports verified facts or reports on political deception)

3.) Journalists takes action and collects more info (asks follow-up questions, finds contradictions in public record, etc.)

In order for a politician to deceive a journalist with their "deception core," then they exploit a couple of failures in the journalist's thinking or process of data collection in order to prevent the third step of collecting additional data -- the politican is trying to get the journalist to accept their framing and not ask questions about missing information.

These following two deception strategies each have three main tactics:

1.) Disrupt or Influence the Journalist's Process of Collecting Data on Deception Core

a.) Block or Remove Deception Core's Attributes

(i.e. "Masking" -- Hiding info by not recording it or disclosing it.)

b.) Diffuse or Confuse Core Attributes

(i.e. "Dazzling" -- Burying info in the footnotes, releasing damaging info on Fridays or on Holidays.)

c.) Divert Attention from the Core

(i.e. "Decoying" -- Red Herring tactics to divert the press' attention.)

2.) Influence the Journalists' Mental Categorization or Interpretation of Deception Core

a.) Deliberately frames and labels core for misclassification

(i.e. "Repackaging" -- Orwellian doublespeak, narrow framing)

b.) Modifies Deception Core so that it is mistaken for something else

(i.e. "Mimicking" -- Creates fictitious events void of substance such as Photo Ops)

c.) Provide weak evidence for the Deception Core so that the Journalist dismisses it.)

(i.e. "Double Play" -- [I haven't figured out how this applies to a journalistic context yet -- Any thoughts?])

Johnson et al cites the terms of masking, dazzling, decoying, repackaging, mimicking, double play from Bowyer '82.

In order for a journalist to counter political deception, then they must be able to read between the lines and find out the motivations and goals of the politician. (Dennett '89, Trope '86, Wellman '90 via Johnston, et al.)

This requires journalists to venture into the invisible interior domains of political psychology in order to induce the intentions behind past behaviors in order to create expectations about future behavior. Without this insight, then the journalist has no alarm mechanism for detecting inconsistencies and countering these six deception tactics.

The basic journalistic detection alarm system would have three conditions for determining whether or not access to information is being blocked or obfuscated by politicians (Johnson et al.)

1.) Some piece of expected political information is missing from the environment or inconsistencies in the expected political behavior are observed.

2.) The absence of this information or inconsistent behavior serves the functional goal of the politician.

3.) Journalist determines that the politician has the capacity to withhold information or create the manipulations that generate the inconsistencies.

If all of these conditions are met, then alarm bells should go off for the journalists that a politician might be trying to hide or manipulate information and they should start digging.

In order to convert these deception detection theories into a functional fraud detection model, then Johnson et al. suggest the following four steps that I have modified in order to to apply to a journalistic context:

1.) Activation: Generate expectations for political statements and behavior based upon previous statements and behavior. Inconsistencies are determined by a large magnitude between the expected and the observed behavior.

2.) Hypothesis Generation: Proposes hypotheses to explain observed inconsistencies -- either intentional manipulation and deception, unintentional errors, or insufficient disclosure.

3.) Hypothesis Evaluation: Assesses the viability of every hypothesis based upon the available confirming or disconfirming evidence.

4.) Global Assessment: Combines all acceptable hypotheses and produces a final diagnostic outcome.

These is the first comprehensive deception detection model that I've really looked at in depth. I'm sure that there are others. I welcome any feedback.

It's important to note that these techniques are domain independent and could be applied to any situation where an adversary is deliberately trying to deceive.

It's also important to note that I discovered Johnson et al's paper from this Mitre Corporation paper that modifies Johnson's deception detection approach in order to apply it to a national security context of intelligence gathering -- The theoretical suggestions from Mitre are just as applicable to journalism, and they include Heuer's Analysis of Competing Hypotheses. I've summarized their suggested modificiations here.