April 2025 – Learning Aloud

Smart Connections finds note connections

Smart Connections discovers and reveals related notes in Obsidian. I started using the Obsidian plugin Smart Connections because I wanted a way to apply AI interrogation of my own notes. I wanted to request cross-note summarizations and generate a variety of sample written products (e.g., blog posts) based on my personal notes and highlights. I ignored an important capability that is claimed based on the product’s name — the identification of note connections.

Without an AI-based method for identifying possible connections among notes, Obsidian relies on the user to establish connections via links and tags. I was aware that other services (e.g., Mem.ai) suggested that a note retention system could do better and offered tags and links, but also made the claim that AI would help surface connections. Some would argue that exploring your Obsidian content repeatedly and finding connections are important parts of the process of personal knowledge management. Constantly working with your notes is an active cognitive activity that encourages connections between what is internally retrievable at a point in time and what you are accessing in Obsidian. New connections first brain to second brain and within Obsidian may emerge. This constant interactive process is suggested by what I would describe as the Zettelkasten practitioners. I don’t think this advice must be rejected for users who want to use AI to surface new connections.

Smart Connections makes use of AI, and the AI creates a numerical representation of the content of each note and stores these as what are called embeddings. You must subscribe to an AI provider via an API, which is far less expensive than a subscription to such a service. You have the option of basing such representations on blocks within notes rather than entire notes. I make use of this option because I store lengthy notes containing book and pdf highlights, such that a representation of an entire note does not represent a level of detail that is very useful for finding something useful in such lengthy notes. In the content that follows, I will show where to turn on block embedding.

Smart Connections works by requesting connections for a note that you have selected. The following image shows Obsidian with Smart Connections active. The green rectangle in the menu bar is used to activate the Connections as opposed to the Chat capability of Smart Connections. The up/down symbol allows you to scroll through the associated notes/blocks from most related to less related. The gear symbol is used to access settings for Smart Connections. The middle panel is the active note, and the right-hand column represents a hierarchy of related notes/blocks.

Getting back to how I think AI may supplement the more hands-on use of Obsidian, I would recommend that in examining connections to a given note that you then use tags or links if you want to create permanent connections.

The extension of Smart Connects from note to note to note to block is worth doing if you do not keep atomic notes. Start with the Gear icon (see image above). This will reveal multiple setting options. What you are searching for are the environment settings. Open these settings with the button shown below.

Once more settings have been revealed, you are looking for Smart Blocks (see below). You turn this option on and specify a minimal length. I did not keep a careful record of the source for advice I followed and I apologize to the author, but I entered 300 characters, and that seems to work well. There are many other settings and I have mostly stayed with the defaults.

Summary:

Smart Connections is an Obsidian plugin (free) that allows AI capabilities to be applied to the notes stored in Obsidian. Chats allows a user to generate AI prompts that are applied to the contents of Obsidian. Connections generates a list of notes (note blocks in the setup I have described) associated with a selected note and is helpful in the identification of such relationships in a large collections of notes.

Writing down a crisis

I began working on a post about the similarities of expressive writing and writing to learn some days ago. In the meantime, the stock market and the general attitude of the country took a decided nosedive. I could not ignore a possible interaction between this dramatic negative mood swing and the original focus of expressive writing so I will just recommend writing to those struggling to deal with our present political crisis and encourage you to read the rest of the post to learn why.

Dr. Jamie Pennebaker found a use for writing that produces both consequential and fairly consistent results. His results relate to clinical psychology which is somewhat outside my own background as an educational psychologist, but I at least can appreciate the impact of an intervention that can be classified as both consequential and fairly consistent as such outcomes are less common than others might imagine when it comes to impacting human behavior. Pennebaker asked college students to think of a traumatic experience from their own lives. His instructions – think about your feelings and emotions related to this experience. I want you to write about this experience for 15 minutes. I will have you do this for three straight days. What you write will be confidential – no one will read what you write. A control group (randomly assigned) was asked to write about their daily routine for the same periods of time. The researchers conducting this study then followed the number of student visits to student health in the following months and found that what Pennebaker eventually described as the expressive writing group had significantly fewer visits. Writing appeared to have an impact on mental health.

I know this seems on the level of magic or weird as I can imagine many reasons this connection might not materialize. Even if the treatment had an immediate consequence on the “problem” why would it follow that the results would be related to medical issues? What if the “problem” was an issue they experienced in their childhood? Why would such a random task during their college days have an impact?

I can’t answer these questions, but hundreds of follow-up studies have produced related results. Pennebaker and other researchers found that expressive writing could enhance immune function, lower blood pressure, reduce muscle tension, and even decrease doctor visits. These benefits were observed across various studies involving participants with both physical illnesses (e.g., arthritis, asthma) and mental health challenges. There has to be something to the benefits of writing.

I first encountered the concept of expressive writing not through my prior work as a psychologist, but because of an interest in the benefits of keeping a notebook. Pennebaker’s work was described in one chapter of Allen’s book “The notebook: A history of thinking on paper”. Once I became interested, I conducted literature searches that might point to an explanation for what about writing might produce this impact. Meta-analytical papers are relevant to the goal of why things work as they do because such papers examine many studies on a given topic, successful and successful experiments, and attempt from this variety of studies to determine what are the factors that contribute to successes and failures. The logic in this approach is that the differences are key to understanding why a technique might be successful and what are the boundary conditions.

The following are the suggested explanations for the benefits of writing.

Catharsis without social risk. You have likely heard of an LBGTQ+ individual “coming out”. This decision when public provides a release from feelings that you have to hide who you are and what you feel. Perhaps expressive writing works in a similar way even though writing is private. This is my example of how catharsis works and I hope this comparison is appropriate.
Cognitive-processing theory. Writing requires concreteness as the abstract and fuzzy ideas in your mind must be made concrete as the ideas are put down on paper. Pennebaker built a digital tool for identifying keywords and concepts in what was written (not in the original study promising anonymity). Those participants with more positive outcomes made greater use of causation words (e.g., because, cause, effect) and insight words (e.g., consider, know) in the content they produced. Perhaps writing helps work out why something happened to you and how significant long term consequences might actually be.
Self-regulation theory. Being able to label stressors and challenges may give the writer a greater sense of understanding and control reducing negative affect leading to greater confidence in better outcomes in the future.

Generative processing as a general explanation for the benefits of writing

I have tried to translate some of these clinical concepts into something more familiar to me. I see similarities in learning and study techniques described as generative learning. In past posts on generative activities, I have explained that the use of a self-imposed or assigned external task encourages productive mental activities. In other words, a learner has the capacity to apply process productively, but for one reason or another does not. The external task (e.g., answering questions, writing summaries, explaining to a peer) encourages these productive thinking behaviors in order to perform the external task and better understanding and retention is produced as a consequence. The cognitive processing of emotional issues may similarly be manipulated by a concrete external task (i.e., expressive writing). This way of thinking seems to fit with the theoretical proposals in the meta-analyses I listed and I think offers a tangible approach that is easier to understand and communicate.

I can’t help thinking about AI as I write this post. How might one encourage tangible “externalization” and processing of life experiences? You may have heard of ELIZA which while not AI could carry on a conversation of a sort through the use of some clever programming that used language patterns built on the input from a user to generate responses and encourage further input on their part. The Wikipedia link in the previous sentence offers more detailed information. Current large language models can now do far more. AI therapy exists and is controversial, but how different is chatting with CHATGPT and writing something you know no one will read?

What about Trump and the stock market? I will write something and put it on Facebook and I do hope someone reads it.

Sources:

Allen, R. (2024). The Notebook: A History of Thinking on Paper. Biblioasis. (Chapter 24)

Frattaroli, J. (2006). Experimental disclosure and its moderators: a meta-analysis. Psychological bulletin, 132(6), 823-865.

Fiorella, L., & Mayer, R. (2016). Eight Ways to Promote Generative Learning. Educational Psychology Review, 28(4), 717-741.

Guo, L. (2023). The delayed, durable effect of expressive writing on depression, anxiety and stress: A meta-analytic review of studies with long-term follow-ups. British Journal of Clinical Psychology, 62, 272–297. https://doi.org/10.1111/bjc.12408

Pennebaker, J. W. (1997). Writing About Emotional Experiences as a Therapeutic Process. Psychological Science, 8(3), 162-166. https://doi.org/10.1111/j.1467-9280.1997.tb00403.

Sharing Notes

I have been retired for several years and miss the social experience of sharing the research process with graduate students and faculty colleagues. Many of my interests are still very similar, but after moving to a different city, I no longer have day-to-day acquaintances with those kinds of interests. This has resulted in an interest in tools that offer note-sharing capabilities.

There is still the challenge of finding others with similar enough interests that sharing is attractive. This post will describe Glasp as a solution. At present this app is free. It offers several ways to share public notes without first having to go through a familiarization process to identify others with related interests, and the type of interaction that follows is up to you. Finally, Glasp has a built-in AI tool that offers an effective way to explore both someone else’s and your own notes.

The following shows Glasp. From left to right, the first column contains personal information and basic controls, the second column lists thumbnails of the pages I have annotated (these are web pages, and other sources such as Kindle books are accessed from an icon on the heading). The final column contains the highlights and notes from a selected source. The drop-down menu is what I wanted readers to understand as it includes a link for locating “like-minded users” (second image). Glasp analyzes the content you have stored and recommends other users with similar interests. I should make clear that you store content as public or private and when you use social capabilities it is only the public content that is visible.

When you select another user, you are provided access to their site (read only) and you select to follow if you are interested.

If you identify other Glasp user by other means, you can simply search for that individual and then access their site to follow that individual.

The AI feature

On your own site or the site of someone you have followed, a button located near the image icon of the site owner, allows access to the AI tool (following image). This is where you submit a prompt to chat with the content on that site (second image).

AI is a good way to explore a collection of content that is unfamiliar to you. With Glasp, you can use the output from a prompt to identify the source notes and move from there to the sources.

Connecting

I am interested in connecting with anyone who finds this tool interesting. You do need to create a Glasp account (free) to follow through. Without adding content to generate recommended matches, you can view my content by entering Mark Grabe in the search box to follow me. Glasp also offers a link that takes you directly to the AI prompt page for a designated user (link to my prompt page). This only connects if you have a Glasp account, but I guess it is provided so you do not have to identify yourself as a follower if an existing user wants to share content in this way.

I like Glasp and would have likely invested more time in the site for my professional reading (mostly journal article PDFs) if I had not already invested many hours highlighting and annotating with a different tool that does not transfer the highlighted content to Glasp. Glasp does include the highlights and notes I have generated from web pages and from Kindle books.