Human-Centric AI: Understanding Technology’s Impact on Society

As a digital anthropologist, studying all digital technology from a human-centric approach allows me to focus on how the technology impacts the humans who use it and human culture more broadly. As I was recently preparing for an invited lecture on AI & Society, I realized that the existing schemes for categorizing AI focused on categorizing the technology, its level of intelligence, and the underlying learning (machine learning, deep learning, nerual networks) and data-gathering mechanisms (sensors, computer vision, etc). But as an anthropologist, I think it’s also important for us to have a human-centered approach to categorizing AI based on how people interact with AI-powered technologies.

These terms aren’t entirely new by any means; we used the term diagnostic AI ” to describe the AI-powered remote heart monitor from my product management days at Preventice Solutions | Boston Scientific and conversational AI to describe the patient engagement chatbot I worked on there. But in the wake of the hype around Gen AI, I’m pulling together a list of human-centered categories to illuminate the current AI landscape and make it more accessible for people who interact with AI on a regular basis, often without realizing. My goal here is to give a common vocabulary without the more technical jargon for people curious about AI and emerging technology.

1. Conversational AI:

• Description: AI systems designed to simulate human-like conversation. They understand and respond to natural language, enabling interactions through voice or text.
• Examples:
  • Virtual Assistants: Siri, Alexa, Google Assistant, Cortana. Used for answering questions, setting reminders, playing music, controlling smart home devices, and more.
  • Chatbots: Used in customer service for answering FAQs, providing support, and guiding users through processes on websites and applications.
  • Voicebots: Similar to chatbots but operate through voice interfaces, often used in call centers for automated customer interactions.

2. Predictive AI:

• Description: AI systems focused on analyzing data to identify patterns and predict future outcomes or trends.
• Examples:
  • Recommendation Systems: Used by e-commerce platforms (like Amazon), streaming services (like Netflix and Spotify), and social media to suggest products, movies, music, or content that users might like based on their past behavior and preferences.
  • Fraud Detection: Used in finance to analyze transaction patterns and identify potentially fraudulent activities.
  • Predictive Maintenance: Used in manufacturing and transportation to analyze sensor data from equipment and predict when maintenance might be needed, preventing costly breakdowns.
  • Demand Forecasting: Used in retail and supply chain management to predict future demand for products, helping with inventory management and resource allocation.
  • Risk Assessment: Used in finance and insurance to assess the likelihood of certain events occurring, such as loan defaults or insurance claims.

3. Automation AI:

• Description: AI systems designed to automate repetitive or complex tasks, often to improve efficiency and reduce human error.
• Examples:
  • Robotic Process Automation (RPA): Software robots that can automate rule-based tasks across various applications, such as data entry, invoice processing, and report generation.
  • Industrial Robots: Used in manufacturing for tasks like welding, assembly, and packaging, often with increasing levels of autonomy and adaptability powered by AI (e.g., vision systems for quality control).
  • Autonomous Vehicles: Cars, trucks, and drones that can navigate and operate with limited or no human intervention, using AI for perception, decision-making, and control.
  • Email Filtering: AI algorithms that automatically sort and filter emails, identifying spam and prioritizing important messages.

4. Perception AI:

• Description: AI systems focused on interpreting and understanding sensory data from the real world, such as images, videos, audio, and sensor readings.
• Examples:
  • Computer Vision: Used for facial recognition, object detection in images and videos (e.g., in security systems or autonomous driving), image classification, and medical image analysis (which overlaps with diagnostic AI but the underlying technology is perception).
  • Speech Recognition: Converting spoken language into text, used in virtual assistants, voice search, and transcription services.
  • Natural Language Understanding (NLU): While part of Conversational AI, NLU as a standalone category focuses on the AI’s ability to comprehend the meaning and intent behind human language, which is crucial for many applications beyond just conversation (e.g., understanding commands in smart devices).
  • Sensor Data Analysis: AI algorithms that process data from various sensors (e.g., temperature, pressure, motion) to understand the state of a system or environment, used in IoT devices and industrial monitoring.

5. Planning and Decision-Making AI:

• Description: AI systems that can analyze complex situations, formulate plans, and make decisions to achieve specific goals. This often involves elements of prediction and automation.
• Examples:
  • Route Optimization Software: Used in logistics and transportation to plan the most efficient routes for delivery vehicles, considering factors like traffic, distance, and delivery windows.
  • Game-Playing AI: AI that can play complex games like chess, Go, and video games at a high level, requiring strategic planning and decision-making.
  • Resource Allocation Systems: AI that can optimize the allocation of resources (e.g., staff scheduling, energy distribution) based on various constraints and objectives.
  • Autonomous Agents: AI systems designed to operate independently in complex environments, making decisions and taking actions to achieve their goals (e.g., in robotics or simulated environments).

6. Generative AI:

• Description: AI systems designed to generate new, original content that resembles the type of data they were trained on. This can include text, images, audio, video, code, and more.
• Examples:
  • Text Generation: Tools that can write articles, poems, scripts, emails, and other forms of text based on prompts or input data (e.g., GPT-4, LaMDA).
  • Image Generation: Tools that can create realistic or artistic images from text descriptions (e.g., DALL-E 2, Midjourney, Stable Diffusion).
  • Audio Generation: Tools that can create music, sound effects, and even synthesize realistic human speech (text-to-speech with lifelike intonation).
  • Video Generation: Emerging tools that can create short videos from text prompts or other input. (e.g., Adobe Firefly, Open AI’s Sora)
  • Code Generation: AI systems that can write computer code in various programming languages based on natural language descriptions or specifications (e.g., GitHub Copilot).
  • 3D Model Generation: AI that can create 3D models of objects and scenes from text or image inputs.

7. Diagnostic AI:

• Description: AI systems focused on analyzing data to identify patterns and determine the cause or nature of a problem, condition, or fault. This is prevalent in healthcare and engineering but can be applied to various domains.
• Examples:
  • Medical Image Analysis: AI algorithms that analyze medical images (X-rays, CT scans, MRIs) to detect diseases or anomalies (e.g., cancer detection in radiology).
  • Analysis of Electronic Health Records (EHRs): AI that processes patient data to identify potential diagnoses, predict disease risk, or suggest further investigations.
  • Fault Diagnosis in Machinery: AI systems that analyze sensor data from equipment to identify malfunctions or predict failures (predictive maintenance).
  • Network Anomaly Detection: AI that monitors network traffic to identify unusual patterns that may indicate security breaches or system problems.
  • Environmental Monitoring: AI used to analyze environmental data to identify pollution sources or ecological imbalances.

8. Surveillance AI:

• Description: AI systems designed to monitor and analyze data from various sources (primarily visual and auditory) to track activities, identify individuals, detect anomalies, and potentially predict future behavior for security, safety, or other monitoring purposes.
• Examples:
  • Facial Recognition Systems: AI that can identify individuals from images or video feeds, used in security, access control, and law enforcement.
  • Object Detection and Tracking in Video: AI that can identify and follow specific objects (e.g., vehicles, people) in video surveillance footage.
  • Behavioral Analysis: AI that analyzes patterns of movement and activity to detect suspicious behavior or potential threats in public spaces or restricted areas.
  • Crowd Monitoring: AI that analyzes video feeds of large gatherings to estimate crowd density, detect unusual movements, or identify potential safety hazards.
  • License Plate Recognition: AI systems that automatically identify vehicle license plates from images or videos.
  • Audio Surveillance Analysis: AI that can analyze audio recordings to detect specific keywords, sounds of distress, or unusual acoustic patterns.

Of course, these AI tools can be combined, yielding powerful results. A recent Nature article talks about the promising potential of a diagnostic AI capable of carrying out a diagnostic dialogue with patients, particularly for improving accessibility and quality of care.

Gen AI have currently captured much of the public imagination and conversations around AI, but AI is so much more. While I’ve been preparing for guest lectures and public talks on AI over the past year, I’ve been thinking a lot about AI and how to categorize AI to fully demonstrate the landscape of AI applications in use today for non-technical audiences. If you’ve made it all the way to the end, I’d love feedback and suggestions.