Understanding the Nuances of Sentiment: How Generative AI Can Help Brands Navigate Public Perception

I. Introduction: The complexities of public sentiment

In today's digital age, understanding public sentiment is more critical than ever, especially during times of crisis. Online conversations are dynamic and multifaceted, reflecting a wide range of emotions, opinions and intentions. Yet, traditional methods of sentiment analysis often fall short of capturing the true complexities of these conversations. These older tools struggle to grasp the nuances of language, context and intent that drive human communication.

For years, social media managers and brand managers have grappled with the frustrations of relying on sentiment analysis tools that provide an incomplete or inaccurate picture of how their audience truly feels. These tools, often based on simplistic positive/negative categorizations, miss the subtleties that shape public perception. However, the emergence of generative AI offers a powerful new approach to understanding and navigating these complexities.

Generative AI – large language models in particular – possess a far greater capacity to analyze the nuances of language, understand the context behind online conversations and ultimately help brands effectively navigate public perception.

The challenge with traditional sentiment analysis tools has been their inability to move beyond basic positive and negative labeling. For example, tools relying on "bag of words" approaches – essentially lists of positive and negative terms – struggle to interpret sarcasm, irony or the subtle shifts in meaning that context creates.

An example sentiment table.

Imagine someone sarcastically posting, "Oh, that was an amazing meal," after a disappointing dining experience. A traditional tool, focused solely on the word "amazing" might misinterpret the sentiment as positive.

Similarly, the use of irony can confound these tools. If someone says, "Well, that sucked," in a context where it's clear they mean the opposite, a simple keyword-based analysis would miss the true sentiment. Even slight changes in wording, such as the addition of a preposition like "the" (e.g., "That was shit" vs. "That was the shit") can completely reverse the meaning, but traditional tools often fail to recognize these crucial distinctions.

These limitations stem from the fact that older natural language processing tools often disregarded crucial elements of human communication. They might strip out stop words like "the" or "a," which can significantly alter meaning. They also struggle with the fact that language evolves rapidly, particularly online and within different generational groups. A Gen Z user might use common words differently than a Gen X user, rendering a static-bag-of-words approach ineffective.

Generative AI, powered by large language models, takes a fundamentally different approach. These models are trained on massive datasets of text enabling them to develop an understanding of language based on statistical association and context. Word order, context and even stop words are all taken into account. This allows generative AI to recognize the difference between "something is shit" and "something is the shit" because it doesn't discard the elements that shape meaning.

As a result, generative AI is significantly better at interpreting sentiment, understanding the emotions behind the words and even inferring user characteristics and motivations. While not perfect (it still struggles with aspects like body language and tone of voice that are absent in text), generative AI offers a major leap forward in sentiment analysis, providing brands with a powerful new tool to understand and respond effectively to the complexities of public perception.

II. Going beyond basic sentiment analysis

As discussed, traditional sentiment analysis tools often fall short in capturing the nuances of human language. Their reliance on simple positive/negative categorization limits their ability to accurately interpret the wide range of emotions and intentions expressed in online conversations.

Generative AI, with its advanced language understanding capabilities, transcends these limitations, offering brands a deeper and more insightful understanding of public sentiment.

One of the key benefits of using generative AI for sentiment analysis is its ability to identify a wider spectrum of emotions. Unlike traditional tools that focus on basic polarity, generative AI can detect emotions like anger, frustration, fear, sadness, joy, excitement and more.

This granular level of emotional understanding allows brands to tailor their responses in a way that is far more specific and effective. For example, if a brand detects a surge of anger and frustration within a specific customer segment, they can craft a response that addresses those specific emotions, acknowledging the concerns and offering solutions.

Generative AI excels at understanding the context and motivations behind different sentiments expressed. It recognizes that the same words can carry different meanings depending on the surrounding text and the broader conversational context. This contextual awareness is crucial for accurately interpreting online conversations, particularly during a crisis when emotions may be heightened and language can be more nuanced. By understanding the context, brands can avoid misinterpretations and craft more appropriate and empathetic responses.

Generative AI's ability to segment audiences based on their specific concerns and emotional states is another powerful advantage. It can analyze vast amounts of data to identify patterns and group individuals with similar sentiments. This segmentation enables brands to tailor their communication strategies for different audience groups, ensuring that messages resonate with the specific needs and concerns of each segment.

The power of generative AI lies in its ability to analyze language in a way that mirrors human understanding. It recognizes that language is not just a collection of individual words but a complex system where meaning is shaped by context, word order and even seemingly insignificant elements like stop words.

By understanding these nuances, generative AI unlocks a deeper understanding of public sentiment, enabling brands to move beyond basic positive/negative categorization and engage with their audiences in a more meaningful and impactful way.

Consider the example of a brand facing a social media crisis related to a data breach. For this, we’ve simulated some angry customer responses on social media to a fictional CPG/FMCG store, MarketMart, helmed by fictional CEO Bob Market.

● @BobAtMarketMart I'm finding you, Bob. And you're gonna pay. 😡 #MarketMartCEO

● @BobAtMarketMart YOU ARE A THIEF! I hope you rot in hell! #MarketMartCEO #DataBreach

● @BobAtMarketMart You better be lawyered up, buddy. We're coming for you! 🔪 #MarketMartCEO #Lawsuit

● @BobAtMarketMart You think you can get away with this, Bob? Think again. We're watching you. 👀 #MarketMartCEO #DataTheft

● @BobAtMarketMart You're a ****ing disgrace! Hiding like a coward. Explain yourself! #MarketMartDataBreach

● @BobAtMarketMart You're a dead man walking, Bob. You messed with the wrong customers. ☠️ #MarketMartCEO #Violence

● @MarketMart I'm freaking out! My bank details are GONE! What the hell am I supposed to do now?! 😭 #MarketMartHack

● @MarketMart I'm never shopping at MarketMart again! BoycottMarketMart! #MarketMartBreach #Boycott

● @MarketMart My identity is STOLEN! I will sue you for everything you're worth! 💰 #MarketMartHacked #Lawsuit

● @MarketMart OMG! I'm so scared! What if they drain my accounts? I'm broke AF as it is! 😭 #MarketMartDataBreach

● @MarketMart This is worse than that time I accidentally bought generic ketchup. WAY WORSE. 🤬 #MarketMartHack #KetchupGate

● @MarketMart YOU HACKED MY LIFE! I DEMAND COMPENSATION! #MarketMartBreach #GiveMeMyMoney

● @MarketMart I knew I shouldn't have trusted a store that sells off-brand Oreos. This is what I get. 🙄 #MarketMartHacked #Regret

● @MarketMart My entire life is online. Now it's in some hacker's hands thanks to YOU. I'm screwed. 😫 #DataBreach

● @MarketMart This is UNACCEPTABLE. You'll be hearing from my lawyer. Prepare to pay! #MarketMartHacked #DataTheft

● @MarketMart WTF?!?! My data?! You gonna pay for this, you corporate hacks! 🤬 #MarketMartBreach #DataBreach

● @MarketMartCares Absolutely disgusted with @MarketMart. No words. Just pure rage. 😡 #MarketMart #NoResponse

● @MarketMartCares FIX THIS ****ING MESS! NOW! #MarketMart #DataBreach

● @MarketMartCares Seriously considering driving to HQ and raising HELL. 🤬 #MarketMart #NoResponse

● @MarketMartCares Someone needs to go to jail for this! #MarketMart #DataBreach #CriminalNegligence

● @MarketMartCares This is a NIGHTMARE! I can't sleep! I'm constantly checking my bank account! 😫 #MarketMart #Anxiety

● @MarketMartCares WHERE IS THE EXPLANATION? WHERE IS THE APOLOGY? WHERE IS MY MONEY?! #MarketMartBreach #NoCommunication

● @MarketMartCares YOU STOLE MY INFO! I'M CALLING MY LAWYER! CLASS ACTION LAWSUIT INCOMING! 😡 #MarketMartHacked

● @MarketMartCares Crickets from @MarketMart. Typical. They don't give a damn about us. 😠 #MarketMartSilence

By using generative AI, they can not only identify the overall sentiment surrounding the issue (e.g., negative) but also pinpoint specific emotions like anger, disappointment and anxiety within different customer segments.

This allows them to craft targeted responses that address the specific concerns and emotional triggers of each segment. They can acknowledge the anger, express empathy for the disappointment and offer clear solutions to alleviate anxiety. This nuanced approach is far more effective than a generic response that fails to address the specific emotions driving the conversation.

In addition, generative AI can help brands identify the motivations behind different sentiments. For example, if a segment of customers is expressing frustration due to a lack of communication from the brand, the AI can identify this as a key driver of negative sentiment.

This allows the brand to proactively address the issue by improving communication channels and providing timely updates. By understanding the "why" behind the sentiment, brands can take targeted actions to address the root cause and improve customer satisfaction.

In this example, generative AI applied a sentiment score, then classified the emotion using Plutchik’s Wheel of Emotions, a framework from psychology and behavioral sciences. After that, it classified each post into a set of pre-defined categories to make analysis easier and faster.

By moving beyond basic sentiment analysis, generative AI empowers brands to gain a deeper understanding of their audience's emotions, concerns and motivations. This understanding is invaluable for navigating public perception, especially during a crisis, and for building stronger relationships with customers based on empathy, authenticity and responsiveness.

III. Sentiment analysis through the lens of the ICP

While understanding overall public sentiment is crucial, brands must also prioritize the sentiment of their most valuable customers — their ideal customer profile (ICP). Generative AI allows brands to analyze public sentiment through the lens of their ICP, providing targeted insights that inform communication strategies and prioritize response efforts.

By filtering and analyzing social media conversations, customer service interactions and other data sources based on the characteristics of your ICP, you gain a deeper understanding of how your most important customers feel about your brand, its products and its actions, particularly during a crisis.

This targeted analysis allows you to do the following:

Prioritize responses based on the sentiment of most valuable customers

During a crisis, it's essential to prioritize response efforts. Generative AI can identify high-value customers who are expressing strong negative sentiment, allowing you to quickly address their concerns and mitigate potential damage to key relationships. For example, if your brand's ICP consists of CMOs and marketing leaders, you can prioritize responding to negative feedback from individuals matching that profile, ensuring that your most important customers feel heard and valued.

Tailor communication to address the specific concerns and emotional triggers of target audience

Knowing how your ICP feels allows you to craft highly targeted and effective messages. Generative AI can identify the specific concerns, emotional triggers and language preferences of your ICP. By understanding the unique needs and anxieties of your ICP, you can create messages that address their specific pain points and offer solutions that resonate with their values and goals.

Ensure that messaging resonates with people who matter most to the brand's success

Ultimately, focusing on the sentiment of your ICP ensures that your messaging resonates with the people who matter most to your brand's success. By understanding how your most valuable customers feel and what they need, you can build stronger relationships, increase loyalty and protect your brand reputation during challenging times.

Consider a brand that has identified "innovation-driven" CMOs in the retail industry as their ICP. During a crisis related to a data breach, they can use generative AI to analyze social media conversations and customer service interactions, filtering for individuals who match the ICP's characteristics.

This allows them to understand how this specific group feels about the breach. Are they expressing anger? Fear? Uncertainty?

By understanding the specific emotional responses of their ICP, the brand can tailor its communication strategy accordingly. For example, if the ICP is expressing concern about the security of their data, the brand can create targeted messages that emphasize the steps taken to address the breach and reinforce their commitment to data security. If the ICP is expressing frustration with the lack of communication, the brand can prioritize providing timely updates and transparent information to this specific group.

This targeted approach not only strengthens relationships with the brand's most valuable customers but also demonstrates that the brand understands and values their specific needs and concerns. It builds trust and reinforces the brand's commitment to customer satisfaction, even during a crisis.

By analyzing sentiment through the lens of the ICP, brands can ensure that their crisis communication is not only effective but also deeply resonates with the individuals who are most critical to their long-term success. It allows them to prioritize their efforts, tailor their messaging and build stronger relationships with the customers who matter most.

IV. Generating response candidates for different sentiment levels

Beyond analyzing sentiment, generative AI can also be used to create a range of response candidates tailored to different levels and types of sentiment expressed. This empowers brands to ensure that their communications are not only appropriate but also effective for the specific emotional context of the situation.

Instead of relying on generic or pre-written responses, the brand can leverage generative AI to create empathetic responses that acknowledge the customers' frustration and validate their feelings.

Generative AI can not only come up with response candidates, but it can also evaluate them against crisis communications best practices and devise responses that meet the needs of the customer.

Consider our previous example of the fictional MarketMart company that has suffered a catastrophic data breach. Here, we use generative AI to design a scoring rubric to evaluate whether a company response is likely to make things better or worse:

If a brand detects a segment of their audience experiencing fear or uncertainty due to a product recall, generative AI can be used to craft reassuring messages that address those specific emotions. These messages might focus on emphasizing the brand's commitment to safety, providing clear information about the recall process, and offering support and resources to help customers navigate the situation.

For audiences seeking clarity and facts during a crisis, generative AI can assist in creating informative updates that provide accurate and timely information. These updates might include details about the situation, the steps the brand is taking to address it and any potential impact on customers. By providing clear and factual information, brands can help reduce uncertainty and speculation, fostering trust and transparency.

Generative AI can assist brands in tailoring their responses to align with the specific communication styles and preferences of their ICP.

For example, if a brand's ICP consists of data-oriented executives, they might prefer responses that are concise, factual and data-driven. Generative AI can be used to create response candidates that meet these specific preferences, ensuring that the brand's communication resonates with their most valuable customers.

It's important to note that while generative AI can be a powerful tool for creating response candidates, it's essential to maintain human oversight in the process. The AI-generated responses should be reviewed and edited by experienced communicators to ensure accuracy, appropriateness and alignment with the brand's voice and values.

By leveraging generative AI to generate response candidates tailored to different sentiment levels, brands can improve the speed, effectiveness and empathy of their crisis communication. This helps build trust, protect brand reputation and strengthen relationships with customers during challenging times.