The Ubik Portals Bible

What follows is the internal design document for Ubik Portals, the classroom side of Ubik: assignment types teachers could shape, and a five-point scale that decided exactly how much AI help each assignment allowed. It is reproduced as it was written, margin notes and all.

Dictionary

HW = Homework // CW = Classwork

Baseline rule across all assignment types: teachers have full transparency of students' use inside UbikAI, with live-time stats and interactive data tools to help them understand students' learning habits.

We believe: giving teachers the tools to build AI experiences per assignment (how they see fit) fosters authentic teacher-student collaboration and high-quality, individualized learning environments for students.

ChatHW: this flow is a student-facing homework redesign. Teachers assign readings with comments and annotations at minimum, paired with an AI chat experience grounded in the assignment's context, custom generative friction, and a time limit made by the teacher. Students don't upload a write-up or any notes. Instead, the dispersed comments throughout the reading, plus a robust back and forth about the reading with the chatbot, lets students discover their interests stress-free as they pertain to the learning goals and fosters an authentic and effective learning environment.

This is a five on the scale of AI assistance, chat-based notes with no friction to promote exploration into topics that interest specific students. Teachers then assess the chat history using a set amount of time spent talking with the agent.

ResearchHW: our ResearchHW assignment type consists of three core steps.

1. First, teachers pick how many topically relevant academic sources students must find.
2. Then, decide how many annotations students must make throughout their findings.
3. Last, teachers customize the research assistance by selecting a number from 0 to 5 (0 being no AI assistance, 5 being full collaborative help) and adding custom rules where they see fit, like "Do not give students answers to questions; question back with exploratory positions that provoke new questions."

Teachers have immense control over our ResearchHW assignment type; they can add custom notes on reading to indicate important key sections, exclude databases, or suggest specific papers. After students find their annotations and academic sources, teachers can require students to do an annotated bibliography with 0 AI assistance.

Teachers can also use this assignment type on personal files or scanned readings. Build research assignments with documents and PDFs that are not in databases but are teacher-owned and syllabus-relevant.

// Research assignments can also be CW assignments, with live time view over students' presence in the portal and engagement with the classwork. //

ReadingHW: reading assignments let teachers turn their PDFs into AI-ready chatbot experiences based solely on the context of the reading. Like in research assignments, teachers decide how much generative friction the chatbot has and any custom rules to help students meet learning goals. ReadingHW can be completely AI-free if the teacher believes the assignment would be more educationally positive without an AI assistant. Paired with our analytic tools, teachers have full transparency and knowledge about the student's time spent on reading, along with many metrics to display learning efficacy and progress. A ReadingHW with full AI collaboration lets students interact and explore the paper with tools like AI-driven quote finding, summarization, AI-posed questions, related article suggestions, highlighting and AI notes, and many more.

ClassHW: teachers can turn any assignment (except QuizHW) into a ClassHW assignment. Collaborative assignments are great for fostering conversation and exploration between students. Teachers can set the required number of comments, replies, and questions students make.

// usually suitable for uploaded readings, educational videos, or fostering discussion //

GroupCW: when it's time for in-class group work, teachers can use UbikAI to automatically and evenly create well-balanced groups with varying performance and learning levels. For example, in a History class, students are broken into groups of 5 and individually tasked with finding different academic sources on the same topic. The History teacher has auto-selected groups and sets each group to find ten sources with a generative friction setting of "2" (use AI to inspire ideas and help refine searches but cannot annotate, write, or summarize), ensuring students are reading through their findings and not using AI to generate answers or directly link them to papers.

When making GroupCW or GroupHW assignments, teachers decide.

WriteHW: WrittenHWs mimic an essential upload submission on any LMS. On Ubik Portals, students can upload directly from their Drive, in-browser text editor, or PDFs. For teachers, all student work gets checked for plagiarism and AI generation. Still, most importantly, our AI synthesizes it to help teachers understand how their students are learning and whether they are meeting the teachers' teaching goals.

// WriteCW is more trackable and transparent with live time stats and public student profile presence, so teachers can see when students are working on the reading class or using tools outside the Ubik Portals suite. Teachers can also set friction rules so the AI assistant does not generate text for use in the final written piece. //

QuizHW: quizzes are used after readings to track students' information retention. This assignment type has 0 AI assistance; teachers can optionally require quizzes after every reading. Teachers can form the questions or let the AI craft individualized questions based on students' specific learning needs.

// Watch and Respond: teachers can link YouTube videos or personal MP4s and set discussion points where students can answer critical questions, leave comments, or ask relevant topical questions directly to the teacher. //

// Refine & Define: teachers assign students to generate an entire essay with the AI assistant in a text editor inside Portals. Then, the AI assistant works backward with the student to edit down and find all the mistakes and unimportant, irrelevant, or poorly written parts of the paper to give examples of what AI can't do. //

Ubik's Generative AI Assistance Scale

Ubik Portals gives teachers a Generative AI Assistance Scale (GAIAS) designed to address educators' challenges with student AI use in and out of the classroom. Based on extensive teacher feedback, our solution allows educators to create customized, interactive AI assignments that set appropriate boundaries for AI usage in school. This flexible system helps standardize AI guidelines within each class, promoting responsible use while enhancing learning. With Ubik Portals, teachers can confidently integrate AI into their curriculum, ensuring students are well-prepared for an AI-driven future. The generative scale has 5 points ranging from no AI tools to full collaboration. Unlike current methods of enforcing and implementing AI scales for homework, the Ubik Portals generative scale has hardline tools associated with the numerical values teachers can pick from. We pair this with added custom guidelines the teacher sets to ensure their views and pedagogy are used and carried throughout the assignment generation and exploration with the students.

Why is a generative AI scale important for students and teachers?

• Implementing a generative AI scale supports thoughtful AI use inside and outside the classroom, helping students develop a solid educational foundation by learning when and how to use AI tools effectively.
• When discussing how students will confirm compliance with the generative AI assistance scale, the author says, "Students must disclose that they used AI and submit a link to interactions with chatbots." This trust in students is forward thinking but naive: as recent high school students and as people who interview and collaborate with teachers in our development, we can very strongly say that this leaves too much onto the students (mainly 14 to 18 years old) without any regulation or guidance. Pseudo-scientific numerical values that represent how to do homework become a guessing game for students.

Scale + assignment type + preset tools

Back end tools used by the AI assistant

The numbers in the matrix refer to this table: twenty-seven tools the assistant could be limited to, written as presets for teachers.

1. Outline generation. AI assistant generates a structured outline for the student paper in research paper format.
2. AI feedback. AI gives students genuine feedback in any context around student answers in relation to learning goals and class rules set by the teacher.
3. AI-suggested papers. When students are searching for papers, AI will recommend related papers either from a portal search engine or filing search engine.
4. Topic generation. When starting an assignment, AI (if prompted) will generate students topic ideas for their assignments based on the assignment information and description set by the teacher.
5. Note analysis. If the student makes a note (a highlight with a point for later, one kind of annotation), the AI assistant will respond with analysis to extend critical thinking on this specific point.
6. Draft refinement. AI assistants will help students craft drafts within the scope of the assignment, meaning: suggest new material to strengthen the argument, adjust length, and help structure the paper to better prove students' research claims.
7. Bibliography creation. Basic bibliography ready for students to export and use within work; teachers can set citation format prior.
8. Annotated bibliography. Annotated bibliography with teacher-set word count for citation summarization, and citation format.
9. Paper summarization. When gathering research (academic sources), AI assistants will summarize papers in academic language and detailed summary with highlighted points for students to explore.
10. Collaborative editing. When editing a written upload assignment, if a generative level of "4" or higher is selected, AI assistance will help students go through drafted work and make highlighted suggestions for fixing grammar, structure, and repetition issues in writing. Only available on already completed versions of a final draft.
11. Grammar check. Basic highlighted grammar suggestions when writing inside the text editor.
12. Human forced grammar. If a generative level of "0" is selected, AI is not allowed to make any suggestions for grammar.
13. Evidence analysis. AI will tell students if evidence from found academic sources is related to the research assignment, and if so, why. AI will also help students piece this information into larger research claims.
14. Draft generation. Draft generation, unlike outline generation, is only available when a student has written analysis and found research for their assignment, after an outline is completed.
15. Annotation feedback. After any student-made annotation inside any found academic papers or teacher-uploaded readings, AI will respond after the annotation is made to provoke further insight into the question, note, or highlight.
16. AI-found annotations. AI will automatically make highlights throughout papers with relevant annotations based on queries by students.
17. Search refinement. AI will help students rewrite search queries, helping find better, more relevant work.
18. Search recursion. AI will try to search through the entire knowledge base for search results.
19. Interactive exploration. When students are completing ChatHW, AI will go through entire papers and prompt students to explore specific points that relate to the topic idea; students are then required to complete set amounts of annotations and time-based conversation with the AI assistant. This helps students critically think about readings and provides in-depth analytics for teachers' understanding.
20. AI-driven suggestions. The AI assistant will make suggestions to students for guidance inside work or when trying to start work. AI can provide students with explanations of steps, help students stay on track, and make sure students are aware when information or topics may be out of line with assignment goals.
21. In-depth questioning. When students are completing research and reading assignments with a generative level of "4" or higher, AI will question students when making annotations or claims with the assistant.
22. Highlighting. AI can make highlights on student-found academic sources or teacher-uploaded trusted sources.
23. Critical feedback. An AI assistant will provide students with insights into their work in any context: an annotation, question, written work, class feedback, learning feedback, general feedback in the tone and candor of the teacher.
24. Idea generation. When starting a research assignment or a written assignment, students can ask the AI assistant to help generate topic ideas, search queries, and papers based on the assignment.
25. Content exploration. When completing a research or reading assignment with a generative level of "5", if a student has completed the assigned amount of annotations on a paper, AI will recommend similar papers based on the completed paper with the assigned amount of annotations.
26. AI-driven hints. If any assignment has a generative level of "2", the AI assistant will only reply to students with suggestions and hints for how to get the answer without the help of AI. When a student is completing a written assignment, AI will suggest how to start writing, or recommend papers related to the topic to help students form claims and analysis. When a student is completing a research or reading HW, the AI assistant will only recommend student-suggested keywords to look for when reading, or critical themes that relate to the assignment, but will not have any highlighting ability.
27. AI-suggested idea expansion. When a student is completing a written assignment with a generative level of "5", if prompted, AI will give students new topics to further help prove points, with recommended papers provided by the assistant.