Beyond Codes: Will Machines Grasp Human Sorrow and Joy?

The rapid rise of AI chatbots has raised ethical concerns, excitement, and employment worries in almost equal measures. But are the stakes about to be upped again?

MUO VIDEO OF THE DAY

SCROLL TO CONTINUE WITH CONTENT

If there is an Achilles heel to these tools, it’s the inability to factor human emotions into replies. However, with advances in the field of “emotional AI,” it’s possible that we are about to witness another huge leap forward in AI technology.

Disclaimer: This post includes affiliate links

If you click on a link and make a purchase, I may receive a commission at no extra cost to you.

An Emotional Problem

Understanding human emotions can be complicated, even for humans. Despite it being something we begin learning at birth, we can still frequently misread another’s emotions. To train machines in a skill that humans haven’t mastered is an enormous challenge.

However, the field of emotion AI, also known as affective computing, is making remarkable strides. To understand how emotional AI works, comparing it to how humans interpret the emotions of others is important. The process can be broken down into three main areas:

• Facial expressions/mannerisms: Somebody beaming like a Cheshire cat is obvious. But what about tears? They could be tears of joy or sadness. Then there are the subtleties and fleeting expressions that we barely notice but give you subconscious clues about others’ emotions.
• Body language: Again, there are lots of clues here that humans use almost subliminally to determine emotional states.
• Voice inflection: The tone and inflection of a voice can be a strong indicator of an emotional state. For example, recognizing the difference between joy and anger often lies in the nuances of how something is said.

The nuances of human emotions are where the challenges arise. To address these challenges, emotion AI uses a range of techniques.

How Does Emotion AI Work?

Similar to how AI chatbots rely on huge databases called large language models (LLMs) to generate responses, emotional AI also relies on a massive dataset. The main difference is the form of the data.

Step 1: Gathering the Data

Emotional AI “models” gather data from a range of sources. Like LLMs, text makes up a part of the model. But emotional AI models also use other forms of data too, these include:

• Voice data: This could be from recorded customer service calls or videos, among other sources.
• Facial expressions: This data can be gathered from a range of sources. One common way is to record volunteers’ expressions through captured phone video.
• Physiological data: Metrics like heart rate and body temperature can be measured to determine the emotional state of volunteer participants.

The collected data can then be used to determine human emotional states. It is worth noting that not all emotional AI models will use the same type of data. For example, a call center will have little use for visual and physiological data. Whereas in healthcare, the inclusion of physiological data is incredibly useful.

Step 2: Emotional Recognition

How data is used to understand emotional states varies depending on its type:

• Text analysis: Techniques like sentiment analysis or natural language processing are used to interpret written text. These can identify keywords, phrases, or patterns that indicate emotional states.
• Voice analysis: Machine learning algorithms analyze aspects of a person’s voice, such as pitch, volume, speed, and tone, to infer emotional states.
• Facial expression analysis: Computer vision and deep learning techniques are used to analyze facial expressions. This can involve recognizing basic expressions (happiness, sadness, anger, surprise, etc.) or more subtle “micro-expressions.”
• Physiological analysis: Some emotional AI systems can analyze physiological data like heart rate and temperature to determine emotional states. This requires specialized sensors and is typically used in research or healthcare.

The specifics of how emotional AI works vary depending on the purpose of the application. However, most emotional AI models will rely on at least one of the listed techniques.

Step 3: Generating a Response

The final step is for the AI model to respond appropriately to its determined emotional state. How this response manifests itself depends on the purpose of the AI. It could be in the form of warning a call center operative that their next caller is upset, or it could be personalizing the content of an app.

The full spectrum of uses for this technology will be massive, and organizations are already putting it to various uses.

What Are the Applications of Emotional AI?

AI, in general, is somewhat of a technological multi-tool, and emotional AI is no different. As the technology develops, the spread of uses will widen considerably, as witnessed by the variety of tasks it is already performing:

• Call centers: Emotion AI is being integrated into call centers to assist agents in identifying the emotional state of customers.
• Advertising: Marketing agencies monitor teams of volunteers to assess their emotional response when viewing a particular advert. This allows them to tweak content to align with the desired emotional response more closely.
• Healthcare: AI is already helping treat mental health conditions . This field of medicine is one where emotional AI could be of huge benefit.
• Education: Education apps can be trained to adjust the course work and overall “learning experience” depending on the emotional condition of the student.
• Automotive industry: This one is in the pipeline, but emotional AI could prove an invaluable driving aid. Current research focuses on developing systems that can detect the driver’s emotional state. It can then take some form of remedial action if the driver is over-tired, stressed, angry, or simply away in a daydream.

This all sounds well and good, but as with all things AI, it is never that straightforward. The ethical and privacy concerns surrounding generative AI are just as applicable, but now we have human emotions thrown into the mix.

Ethical and Privacy Concerns of Emotional AI

For every benefit that AI brings us—and there are many—there seems to be a corresponding ethical or privacy concern. This innovative technology is operating at the edge of technological know-how. It is also operating at the edge of societal know-how.

The intersection of emotion and technology is littered with complex challenges that need to be addressed if AI is to be a boon and not a burden. Some of the concerns that are immediately apparent include:

• Data privacy concerns: Already a grey area in AI, the inclusion of sensitive emotional data has raised the bar.
• Accuracy: AI chatbots are many things, but their answers are often wide of the mark. The same errors made by emotional AI models can have serious consequences if they occur in applications like healthcare.
• Emotional manipulation: Scammers could use emotional AI to play on people’s feelings with malicious intent.

These concerns are genuine, and a concerted effort to address them is the key to unlocking the full benefits of emotional AI.

Don’t Know Whether to Laugh or Cry

This is a promising technology with huge potential benefits. However, it does carry some “emotional baggage” along in its slipstream. The upside is the huge range of potential applications where this could make a huge difference. Everything from healthcare to more immersive gaming experiences can benefit from emotional AI.

But there are some hefty issues to be dealt with if we are to use this to benefit and not hinder humanity.

SCROLL TO CONTINUE WITH CONTENT

If there is an Achilles heel to these tools, it’s the inability to factor human emotions into replies. However, with advances in the field of “emotional AI,” it’s possible that we are about to witness another huge leap forward in AI technology.

Also read: