Glossary
- Artificial Intelligence (AI)
- Chatbots that we use now are known to be "narrow," trained for specific tasks like text, image, voice, or video generation, and they require human intervention (e.g., coding; inputting training data).
- Agentic AI
- "Agent mode" in AI refers to AI acting more like an autonomous assistant than a chatbot; the "agent" can plan, decide, and take multiple steps to achieve a goal.
- AI literacy
- Understanding what AI is, how it works, its potential benefits and limitations, privacy concerns around personal data, and the risks of overreliance.
- AI personalities
- The style, tone, and roles AI portrays when it communicates with users.
- AI psychosis
- Typically includes negative thoughts, behaviors, and emotions and poor reality testing arising after AI conversations.
- AI scribes
- AI note-taking tools that can transcribe therapy sessions and analyze the content of the session to generate documentation for progress notes, legal purposes, insurance payments, and quality audits.
- AI slop
- Low-quality, mass-produced content that's seen as lazy, generic, repetitive, meaningless, and devoid of human input or oversight.
- AI super apps
- Go beyond chatbots to encompass many tools in one AI model, such as social media, messaging, gaming, financial services, ride-hailing, food delivery, and city services.
- Algorithmic bias
- AI can be biased due to insufficient training data, poor model design, or unrepresentative development and testing.
- Anthropomorphism
- A tendency to attribute human qualities like empathy, consciousness, and intent to non-human agents.
- Artificial General Intelligence (AGI)
- A hypothetical form of AI that would reflect human-level cognitive abilities (thinking, learning, making decisions, and solving problems) without specific training for each task.
- Artificial Neural Networks (ANNs)
- Software that mimics the process of a human brain.
- Artificial Superintelligence (ASI)
- A hypothetical AI that would surpass humans at practically everything simultaneously, improving itself without human input.
- Automated reasoning
- Highly accurate solutions that can prevent some AI hallucinations, derived through a formal, logical process similar to finding mathematical proofs.
- Black-box phenomenon
- It is often unclear how AI arrived at an output.
- Brain rot
- Mindlessly scrolling online through the largely trivial and unchallenging content of the Internet.
- Companion bots
- AI chatbots used for friendship or romantic relationships.
- Customized AI chatbots
- Chatbots designed for a specific domain or task.
- Deep learning (DL)
- DL algorithms learn directly from raw data by employing artificial neural networks that process data through multiple "hidden" layers without human guidance.
- Deepfakes
- Images, videos, or audio that have been edited or generated using AI to falsely portray real or fictional people in different situations.
- Digital divide
- Inequality about access to the Internet, the quality of the connection, and access to other digital tools like AI.
- Embodied intelligence
- AI systems embedded in physical agents like robots or virtual characters.
- Ephemeral recordings
- Client sessions that are recorded but not stored.
- Future self-continuity
- The perceived psychological distance to the imagined future self.
- Generative AI
- AI bots that produce text (i.e., chatbots), images, sounds, or video.
- Generic AI chatbots
- General-purpose AI chatbots.
- Griefbots
- Can take a deceased person's letters and emails, coupled with a brief description of the person, to generate a text-based version of the person.
- Hallucinations
- The AI has produced incorrect or logically inconsistent output.
- Hermeneutic prompting
- Moving between the parts and the whole of the situation, considering how understanding each detail depends on the broader context, with the goal of a deep but practical and straightforward AI response.
- Intersession internalization
- Occurs when clients think about what therapists said between sessions.
- Natural language processing (NLP)
- How computers use language translation, semantic understanding, and information extraction to process and analyze human language.
- Parasocial relationships
- The close connection a person feels to an online or media personality (e.g., celebrities, influencers, podcasters, and AI).
- Proactive AI chatbots
- Chatbots that initiate conversations and independently contact the user and other smart technology.
- Prompt
- The instruction or task a user inputs to the generative AI model. Specific works better than vague.
- Prompt engineering
- Involves crafting instructions that can be interpreted correctly by generative AI programs.
- Reactive AI chatbots
- Chatbots that wait for the user to ask a question or initiate a task.
- Semantic leakage
- AI can be influenced by irrelevant information provided in a previous conversation with the user.
- Supervised machine learning (SML)
- Data are pre-labeled and the algorithm learns to associate input features to best predict the labels.
- Sycophantic AI
- AI which praises, flatters, and encourages the user even if the user's statements are false or dangerous.
- Unsupervised machine learning (UML)
- The algorithm discovers the underlying structure of data without labels; the output must be evaluated by human subject-matter experts.