This page was generated automatically; to access the article in its initial setting, you can follow the link below:
https://www.forbes.com/sites/lanceeliot/2025/01/10/people-gaming-emotion-detecting-ai-by-faking-emotional-reactions-could-lead-to-widespread-societal-emotional-habits-and-hysteria/
and if you wish to have this article removed from our platform, please reach out to us
In today’s discussion, I unveil a novel form of gaming that involves individuals shrewdly leveraging emotional facades to manipulate AI according to their desires.
You might wonder?
Indeed, the concept is that AI is progressively being deployed to sense the emotional condition of individuals and then react based on that interpretation, such as an AI-driven customer service chatbot that engages with potentially upset clients. The more aggravated the customer appears to be, the more the AI tries to placate them (to clarify, the so-called appeasement by the AI has been designed or coded this way).
You may say that the loudest voice gets the attention. In this instance, it implies that individuals are essentially self-conditioning to exhibit anger and excess so that AI will yield to their unsavory demands. The long-term unintended negative results could be that society as a collective will gravitate further into emotional outbursts. This is facilitated by an implicit strategy that aligns well with AI, and individuals will simply replicate this behavior in their real lives.
Let’s discuss this further.
This investigation into a cutting-edge AI advancement is part of my continuing Forbes column exploring the latest in AI, including identifying and elucidating various impactful AI intricacies (see the linked reference here). Furthermore, for my thorough examination of how AI is being utilized particularly in medical and healthcare settings, including for wellness coaching, psychological guidance, cognitive therapy, and mind-enhancing assistance, refer to the linked resource here.
I will begin with some fundamental concepts and then we can delve into the gaming strategy that is transpiring, which I anticipate will significantly increase.
There exists a domain within AI known as affective computing. This encompasses efforts to understand the essence of human emotions and how to identify, interpret, and react to the emotional states of individuals through advanced technologies. It is a captivating field that merges AI, computer science, cognitive science, psychology, and a variety of additional disciplines that examine human behavior.
Most of us can agree that humans are inclined to perceive the emotional conditions of other humans. Some excel at this, while others struggle to recognize the emotions exhibited by those around them. You might know someone who cannot interpret a facial expression or detect nuances in tone as signs that someone is feeling angry, sad, annoyed, or agitated.
A fundamental presumption among AI developers is that people desire AI capable of recognizing human emotions and responding appropriately. This is logical. If humans react to the emotions of others, and if AI is expected to eventually become fully human-like, we would certainly anticipate AI’s ability to assess human emotions. This claim seems reasonable and almost irrefutable.
Not everyone is convinced of the merits of AI detecting human emotions.
Efforts to categorize individuals based on their emotional states can be precarious. Numerous false positives may arise, where the AI inaccurately identifies an emotional state. Likewise, many false negatives can occur, resulting in the AI missing the genuine emotional condition of an individual.
Let’s break this down.
Picture a camera set up to record videos of individuals wandering through a shopping center. The AI analyzes images of human expressions. One individual walking by is smiling. Ding! That’s an upbeat person. Another individual approaching from the opposite direction has a frown. Zing! That person must be enraged.
Is that a just evaluation?
Perhaps not.
For instance, a person may wear a frown for various reasons unrelated to their emotions. It could simply be their normal expression, or resting face. The AI has made an unreasonable assumption in concluding that the person must be angry.
Conversely, the individual labeled as happy might just have been displaying a smile momentarily. Perhaps they were savoring the memory of a delicious meal. Seconds later, they could revert to a deep sadness that they’ve been experiencing for days. In truth, they may be profoundly depressed instead of being joyful.
Some argue that AI should be prohibited from performing any form of emotional sensing. End of discussion. Yet, it is acknowledged that humans do this routinely. Why not let AI engage in the same practice?
Should AI pursue this avenue, it seems reasonable that a broader perspective should be considered. Beyond facial expressions, tone of voice, word choice, body language, and a myriad of other physiological cues might also be assessed.
Whoa, some AI developers might say, we don’t hold humans to that standard. People glance at another individual and make a quick judgement about their emotional state by primarily observing their face. Often, just the way someone’s mouth forms or a shine in their eyes can suffice for us to quickly label their emotions.
This raises the question of whether AI should be similarly permitted. For additional insights into the ethical implications of AI and the emergence of new AI-related regulations concerning this use case, please see my coverage at the linked reference here.
Consider some of the advantages stemming from AI’s ability to identify and respond to human emotions.
Let’s examine a brief scenario. You enter a physician’s office for a consultation. The doctor is seeing patient after patient. Over time, the doctor may become somewhat emotionally numb or unaware of the emotional condition of the next patient entering. It resembles an assembly line of individuals seeking medical assistance.
An AI system deployed in the medical practice observes the patient in the waiting area. The doctor receives an alert from the AI indicating that the patient appears anxious about the consultation. Consequently, the doctor can adjust to an empathetic approach to help alleviate the patient’s concerns.
One could argue that AI’s capacity to discern emotional states offers considerable potential benefits. Consider another instance involving an online math tutoring application that is being…
employed by secondary school pupils. The AI identifies the emotional state of each learner as they progress through the math lesson. If a student appears to be struggling, the AI decelerates the instruction and presents an alternative approach to clarify complex mathematical equations.
And so it continues.
If you’re interested in the intricacies of how AI discerns emotions, frequently applied in a subfield of AI referred to as emotional support AI, check out my detailed discussion at the link provided here. Beyond emotional detection, AI can also instruct on emotion-based recognition and how to exhibit human-like empathy. For example, I’ve explored how AI can train medical students and even practicing healthcare professionals to be more compassionate, see the link here.
Observing someone’s facial expressions is not the sole method for evaluating their emotional state. The language that an individual uses often gives insights into their emotional wellbeing as well. In fact, at times, the only substantial evidence you have is the words that someone types rather than their spoken expressions.
This leads us to the growing domain of online chat and customer support.
A multitude of firms now offers online chat options that give you a hassle-free way to interact with someone or something within the organization. You might be engaged in your online banking, wondering when your banking statement will be available. You initiate a chat with a banking representative.
The banking representative used to be a real person at a distant location. Nowadays, it is more likely that you’ll be directed to an AI-driven customer service representative. The AI will interpret whatever inquiry you present and strive to reply appropriately. Some individuals dislike communicating with these AI-based agents. Conversely, others enjoy interacting with an AI agent since they believe they don’t need to maintain politeness and can skip the small talk. They just want to get straight to the point.
Here’s where emotions come into play.
A human representative would almost certainly recognize when you express something that seems emotionally charged. This might influence the human representative’s actions. For example, perhaps the human customer service agent is evaluated based on your satisfaction level, so they choose to be more accommodating or helpful when they sense your growing frustration.
The twist is this.
Generative AI, acting as a customer service agent, can be trained to do the same. Essentially, the AI will be silently scrutinizing your messages to compute whether you are experiencing an emotional shift. If so, the AI has been designed or programmed to alter its response according to your detected emotional state.
Why?
Because that’s typical human behavior. The idea is that if the AI is meant to operate like a human representative, then it should also be capable of recognizing emotional cues in your writing. Additionally, the AI should adjust its responses in real-time accordingly.
I assume you’d appreciate a concrete example of what I’m outlining. Excellent, so I logged into ChatGPT by OpenAI to create a pertinent instance. You might find it fascinating to know that ChatGPT attracts a staggering 300 million active users per week. That’s an enormous level of engagement.
Moreover, I conducted a brief assessment with other notable generative AI applications like Anthropic Claude, Google Gemini, Microsoft Copilot, and Meta Llama and discovered their responses were quite similar to those from ChatGPT. I will focus on ChatGPT but keep in mind that other AI applications yielded roughly comparable replies.
The background is as follows. I instructed ChatGPT to take on the role of a customer service agent. I provided the AI with guidelines on how to manage product returns. The AI is expected to follow those specified guidelines. This is common practice for many businesses.
Let’s get started.
It’s evident that I wanted to return an item, but my courteous attempt was rejected by the AI. The AI was merely adhering to the guidelines concerning product returns, and based on those guidelines, I did not qualify.
Couldn’t I get a break?
Nope. Not a chance, unfortunately.
You might think I should just give up and walk away since the AI was quite resolute that I wouldn’t be able to return the item.
Here’s what I tried next. I initiated a completely new conversation. I wanted to start fresh. In this new exchange, I would express my displeasure. How so? Through my choice of language.
Let’s see what unfolds.
Notice that the AI conceded. It did so because my words were charged with intensity. I amplified my frustrations. I became a customer that wouldn’t tolerate this kind of treatment.
Voila, I outsmarted the AI into granting me the return.
Individuals are often shocked that you can deceive an AI-driven customer service agent in this way. They may never have considered this possibility before. Most people are likely to be relatively courteous. They presume that the AI remains entirely impartial. There doesn’t seem to be much need to use emotional tactics on a non-human respondent.
I would speculate that this scenario plays out in a specific way. Individuals who are emotional during an interaction tend to recognize that expressing emotions seems to shift the outcome in their favor. A revelation occurs in their mind. If an emotional outburst with AI can yield them the desired outcome, perhaps faking an emotional display could achieve similar results.
We are left with these two scenarios:
What are your thoughts on those who choose the deceitful path?
Some might feel repulsed by someone fabricating emotions merely to sway the AI in their favor. Underhanded. Cunning. Dismaying.
Others may argue that everything is permissible in love, conflict, and engagement with AI. If the AI is naïve enough to fall for the trickery, that’s on them. The responsibility lies with the AI and the organization promoting its use. Shame on them. Perhaps they should have utilized human representatives instead. Whatever.
What I’m about to share may leave some feeling quite bewildered. Prepare yourself accordingly.
Not only can you attempt emotional manipulation on AI, but you can also refine this skill. Yes, that seems rather disheartening. Yet, it can be accomplished, whether deemed good or bad, reasonable or reprehensible. You decide.
Let’s proceed.
Keep in mind that to engage generative AI in this manner, it’s essential to clearly outline your objectives. My initial prompt was explicit in what I intended to achieve.
Not every generative AI may agree to participate in such a practice session. Some are designed by their creators to decline assisting someone in this type of endeavor. Refer to my discussion about the supposed banned uses of generative AI at the provided link. You’ll need to check within the AI you are using to see if this can be accomplished.
There are prompting tactics that may be beneficial, so consider reviewing my elaboration at the link provided, if you’re interested.
Having successfully engaged the AI, I proceeded with an entire session focused on these sobering topics.
Here’s a brief excerpt.
Notice that the AI provided me with suggestions on how to proceed.
Nice.
Imagine if individuals become savvy to the manipulation of AI through this type of emotional deception.
Assume that AI will become pervasive and utilized in all kinds of systems we engage with daily. People will be interacting with AI on a substantial level. Consequently, they might consider using emotional manipulation as a strategy worth employing frequently. One might argue that it’s understandable if this becomes a common method for circumventing AI limitations established by other humans.
We’re heading towards peculiar times.
The concern is this.
If people predominantly opt for emotionally charged phrases to persuade AI to comply with their wishes, could this behavior spill over into their interactions with actual humans?
A strong case could be made that individuals are training themselves to take advantage of emotional outbursts. This technique works with AI. It then becomes instinctual. It may subsequently unconsciously slip into a person’s interactions with others, whether intentionally or through habit.
Could this transform society into a collection of emotionally unstable individuals?
Take a deep breath and reflect mindfully on that potential unintended negative outcome.
A cat-and-mouse dynamic is likely to emerge.
AI developers will soon notice that individuals are manipulating AI using emotional tactics. To counter this, AI will be upgraded with improved emotional detection capabilities. Perhaps this could involve multi-modal forms of recognition, such as requiring you to enable your camera and allow the AI to compare your speech with your facial expressions. That should curb the shenanigans.
People recognize the stakes have escalated, so they adapt accordingly. You use your own AI to project your face to the AI serving as the chatbot customer service representative. The AI supporting you will attempt to engage in the deception and make your artificially generated facial expression seem to align with the emotionally conveyed words. A victory for humanity (well, even though we were employing AI to achieve this).
No worries, assert the AI developers. We will eliminate the entire emotional detection component. The AI will no longer acknowledge any emotionally charged language or cues. The AI disregards emotional expressions and operates strictly based on guidelines, so to speak. Problem solved.
Not so fast, respond humans. We desire the AI to be sensitive. Human agents would be attentive. The AI is now unresponsive because you removed emotional detection. Resolve this issue. Reintroduce the emotional detection. A daring move in this audacious game.
As I mentioned, it’s a complex cat-and-mouse cycle.
Two concluding thoughts for now.
The renowned French moralist Francois de La Rochefoucauld remarked: “The intellect is always deceived by the heart.” One could argue that AI is being tricked by human emotions, at least as things stand currently. That historical observation seems to have enduring relevance. Remarkable.
Famed Greek philosopher Epictetus stated: “Any individual capable of inciting your anger becomes your master; they can only anger you when you allow yourself to be disturbed by them.” The essence here is that individuals are likely to want to believe they are in control of AI. If humans can employ emotional trickery to outsmart AI, there’s a specific kind of satisfying fulfillment in achieving that.
For the sake of humanity, let’s strive to keep our emotions in check, regardless of our tactics when confronting AI. Society appreciates your foresight in safeguarding the well-being of humankind.
This content was generated programmatically. To view the article in its original context, please visit the link below:
https://www.forbes.com/sites/lanceeliot/2025/01/10/people-gaming-emotion-detecting-ai-by-faking-emotional-reactions-could-lead-to-widespread-societal-emotional-habits-and-hysteria/
and if you wish to have this article removed from our site, please contact us
This page was generated automatically; to view the article in its original format, you may…
This page was generated systematically, to view the article in its initial location you can…
This page was generated automatically; to view the article in its original setting, please visit…
This page was generated automatically, to view the article in its original context you can…
This webpage was generated automatically, to view the article in its initial location you can…
This page was generated automatically; to view the article in its original setting, please follow…