Summary.    Can generative AI tools like ChatGPT help us communicate more effectively — and more empathically? Successful communication often relies on speaking to personality traits, cognitive styles, and identity-linked worldviews. With the right prompts, and familiarity...

Few of us are ever successful in changing minds and behaviors by stacking up piles of evidence wrapped in clear communications that support our goals or views. This is most visible in national issues — ranging from climate change to vaccines — but it’s also true of internal corporate challenges. Leaders fail to understand and account for what makes our audience tick and don’t adapt our communications to match their inner wiring and worldview.

Each of us comes pre-wired for how we’ll respond to every issue, situation, policy, and decision. It is never a tabula rasa of rationality. At Ogilvy Behavioral Science, we like to call this tangled, interconnected wiring of personality traits, cognitive styles, and identity-linked worldviews the “Hidden Who.” A wrong word choice or inappropriate framing can create instant rifts — no matter how logical or well-crafted the messaging seems. On the other hand, the right language can make a message click like a key in a lock. Behavioral science studies have shown that empathetic resonance can succeed where attempts to convert will backfire.

Marketers and communications strategists deploy audience segmentation plans all the time, but rarely based on decoding the “Hidden Who.” Admittedly, applying this lens to every communication is a tall order. But new technology may be able to help. Notably, artificial intelligence tools such as OpenAI’s ChatGPT can be a very effective empathy assistant. In my experiments, I found that ChatGPT can create audience-tailored versions of communications that may resonate more and clash less with each group. Particularly eyebrow-raising is its ability to do so responding to prompts with behavioral science cues — crafting surprisingly appropriate messages for particular personality traits, worldviews, and so on.

CEOs, leaders, and managers routinely need to communicate their plans, strategies, initiatives, changes, or transformation visions but resort to a least common denominator of conventional wisdom. Instead, it is possible to tailor the communications to the audience or audience segments. An AI such as ChatGPT can help by being prompted to write differentiated messages for each audience group depending on their respective worldviews, internal makeup, and the other invisible factors that affect how a message or intervention is received. They can also simulate conversations to help think through how to deliver a message, generate audience-specific analogies, and write drafts of communications, each customized to the underlying individual or group traits. Consider having an entire toolbox of adapters to fit each audience you’re trying to reach.

Thinking About the “Hidden Who”

To arrive at a deeper understanding of individuals, we have to move beyond blunt measures such as demographics. For decades, behavioral scientists have been doing just that. Here are some ways to decode people that warrant consideration. Ideally, you would test the audiences you hope to reach and persuade, but if you cannot, you can still craft versions of messaging that stand a better chance of resonating with each target subgroup than a generic, single message for all.

Loose Culture vs. Tight Culture

The work of behavioral scientist Michele Gelfand has shown us that companies as a sub-culture can exhibit loose cultures, tight cultures, or a mix. In a tight culture, the perceived norms are clear and the tolerance for deviating from them is low. In a loose culture, all kinds of people, values, and behaviors are either tolerated or embraced. How you speak to a loose culture should be very different from how you speak to a tight culture.

Psychological Capital

So-called “psy-cap” is built on the four H.E.R.O. pillars: hope, efficacy, resilience, and optimism. Your employee base may be high or low on each factor, and that strongly influences their attitudes and behaviors. For example, a very high-hope culture that is very low on efficacy wants to believe but doubts its ability to succeed.

Framing

Framing refers to how a choice is portrayed. Behavioral scientists say that we do not choose between two options so much as how those options are described. A gain frame offers the possibility of a good outcome if you do something. A loss frame warns of a bad outcome if you do not do it. Studies show correlations between an individual’s regulatory focus and which framing is more persuasive.

Regulatory Focus

All humans come pre-wired as being more skewed toward a promotion focus or a prevention focus. Those with a promotion focus play to win, even if it means failure, and tend to be moved by a gain frame. Meanwhile, employees with a prevention focus play not to lose, and prefer to stick with what they know in the face of uncertainty. They tend to respond to a loss frame. Corporate cultures can skew more toward promotion or prevention as well.

Locus of Control

People act according to how much power they think they have over their own lives. Those who believe that chance, fate, luck, God, or powerful others control the outcomes in their lives have an external locus of control. Meanwhile those who believe all outcomes depend mostly on their own actions and choices have an internal locus of control. Motivating and convincing “externals” to act requires overcoming this lack of self-agency.

Personality Traits

The science of personality traits has survived decades of peer-reviewed studies across cultures for decades. It deconstructs human personalities into five separate factors: openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. If you know someone is very low on agreeableness, you will address them with concern for their inherent skepticism and brusqueness. If you know someone is very low on openness, do not talk to them about wild new experiences for motivation.

Worldview

As we see in the ongoing culture wars and spiking partisanship, messenger beats message every time. “Cultural cognition” is a useful framework for understanding these identity-tied, in-group worldviews. It maps individuals or groups on a two-by-two matrix according to certain ways of seeing the world: hierarchical vs. egalitarian, and individualist vs. communitarian. The resulting IDs should sound familiar:

• Hierarchical-Individualists believe in free markets and ruthless competition, and value what they believe to be simpler, clearer traditions and social values of the past.
• Egalitarian-Communitarians feel markets exploit people and ruin the environment, prioritize the greater good, and are motivated to erase inequities.
• Hierarchical-Communitarians prize deference to authority and traditions, but prioritize the whole group over individuals (think military ethos).
• Egalitarian-Individualists are libertarians who cherish their individual liberties, object to outside orders or interventions, and believe in a “live and let live” approach.

How Machines Can Help Us Communicate

Already, we have seen how prompting ChatGPT to take on a persona can lead to it “talking” in different voices. (For instance, ask ChatGPT to write a company announcement in the voice of Homer, and then rewrite it in the voice of Homer Simpson.) A more useful task, however, may be to prompt ChatGPT to create message versions that feel comfortable and sensible to very specific audience segments. For instance, if you want to speak with empathy to those who reject any form of health mandates because of concerns about abridging individual liberties. Or taking an uncertainty-avoidant employee base into a new realm of innovation and experimentation.

Artificial intelligence experts caution us that tools such as ChatGPT are not sentient, nor are they search engines or digitized experts. They are, however, increasingly adept mimics. This type of AI — called a large language model — is “trained” by feeding it billions of documents and unstructured information, which it scans for patterns. It then creates responses based both on what it has been trained on and what it projects to be most likely a good response. That makes it good at copying style — it picks up the rhetorical tics and patterns of habits of an author, a group, or scientific studies and findings.

To get started, here are three areas where you can deploy ChatGPT right away:

Versioning

Behavioral science studies have shown that tailoring the language, tonality, and framing to a specific audience who is wired a particular way boosts the effectiveness of that message. ChatGPT can be used to do a form of segmentation and versioning.

Simulating Dialogues

Many times we can benefit from a trial run of interactions or conversations. By prompting ChatGPT with a goal and different audience profiles, it can write dialogues to help aid and instruct us on what to say and how to say it. For example, a manager can simulate a difficult conversation with an employee, prompting ChatGPT with a personality profile of that employee and the context in which the conversation takes place (disciplinary, promotion or raise request, underperformance, etc). Or a doctor could use it to prepare for speaking with a patient who has real concerns about a procedure or even vaccinating their child.

Crafting Relevant Analogies

Many technical, financial, or scientific concepts need simple, clear, relevant analogies to convey meaning. ChatGPT can be prompted with the challenge and create several analogies that may work. Additionally, it can be told to create different analogies for different audiences. For example: “Explain how vaccines work using easy-to-understand, clear, and jargon-free analogies for each of four different audiences: a well-educated parent; a parent with less than a high school education; a teenager; a six-year-old child.”

Putting Prompts Into Practice

Here are four steps to putting ChatGPT into practice in versioning, simulating dialogues, and crafting relevant analogies:

1. Prompt the ChatGPT to incorporate relevant behavioral science principles (see list and descriptions of the “Hidden Who”).
2. Prompt it to identify some opposites within each principle as audience segments, such as high or low versions of a personality trait, or prevention versus promotion focus.
3. Ask it to craft different versions of the same goal-related messaging to best resonate with each of these profiles.
4. Tell it the length and format (letter, dialogue, etc.).

Consider the following prompts:

Using the behavioral science principle of REGULATORY FOCUS assume two separate audiences of PROMOTION and PREVENTION focus. Imagine you are a family doctor who is trying to encourage your patients to vaccinate their children from routine but harmful infectious diseases such as the measles, mumps, etc. Write two versions of a substantive dialogue between the FAMILY DOCTOR and the PATIENT— one dialogue each for each version of a fictional patient. The first patient has a prevention focus. The second patient has a promotion focus. Include the concerns and objections of each patient.

Review the behavioral principle called “psychological capital” before writing two analogies from the fictional CEO of a [describe company] explaining digital transformation as part of announcing their plan for the company. Each analogy will be customized to best resonate with its respective audience. The two audiences are: AUDIENCE 1 is very low “efficacy” and AUDIENCE 2 is low on “optimism.” Make each explanation approximately 300 words.

Review the behavioral framework called “Cultural Cognition.” Write four letters from a fictional CEO to their corporate employees. The letters should explain why the CEO has decided to embrace a new sustainability initiative for the company. Each letter should be customized to best resonate with each of four different audiences according to the cultural cognition framework. Those audiences are: HIERARCHICAL-INDIVIDUALISTS, HIERARCHICAL-COMMUNITARIANS, EGALITARIAN-INDIVIDUALISTS, and EGALITARIAN-COMMUNITARIANS. Make each letter approximately 800 words.

Consider ChatGPT to be a junior coauthor. You will still need to double-check certain aspects and never accept its citations or sources (it has a bizarre practice of fabricating sources and references but then saying it understands that fabricating references is unacceptable). Yet, it can be very useful in crafting first-draft versions of the same messaging for very different audiences, and in helping you practice by writing a simulated dialogue, and crafting specific analogies to convey meaning to different groups depending on each group’s frame of reference. In short, an AI assistant such as ChatGPT can render your communications more effective by resonating better with the Hidden Who in all of us.