OSCAR --Outcome-based System for Consequence Analysis & Recommendations

OSCAR --Outcome-based System for Consequence Analysis & Recommendations


The following is a long term project which I hope will assist me in seeing a longer run consequences of my actions in a data-driven way.

  1. The project is complex and will likely change in scope.
  2. While I have machine learning experience, having another contributor would be both enjoyable and motivating. Please see the bottom of this page for contact info
  3. If released the tool will be open source


OSCAR will aim to generate a consequence score [-1,1] associated with actions.


Due to the inherent complexity of long-run effects and confounding factors, the follow limitations are placed on any v0.x of the system:

  1. Any predicted outcomes will be relative to a specific goal.
  2. Goal attainment is not predicted. The effect, positive or negative, of a given action on the goal is predicted.
  3. There will be a confidence interval associated with any score.



  1. Action Input: the user submits an ‘action’, in JSON, to the interface.
  2. Output: a bounded range score. The sign is the anticipated goal attainment effect direction and the magnitude is the effect size relative to other previously provided scores for the goal
  3. Cluster View: given a goal, a dimensional projection of actions is produced with decision boundary which is partitions actions by score s> 0, s<0. This provides a way of viewing action classes (behaviors) which tend to contribute or detract from attainment.

OSCAR to use an optimized personal dataset, text based sentiment analysis, for training a clustering algorithm to determine predicted outcomes and associated confidence intervals for given decisions and actions.


After model training , the interface and results, currently concieved, will be via a plugin for Obsidian.md

System Structure (current thinking)

  1. Optimized personal data of actions , outcomes of goals, as well as known behavioral patterns associated with positive and negative outcomes will be used for the training dataset.
  2. A logistic regression approach will help classify outcomes.
  3. Causal Inference and DAGs will assist in forming pairs (action, outcome).
  4. A classification algorithm will determine clusters of action categories based on outcome.
  5. Characteristics of these clusters will be checked in a user action
  6. Nearness to a decision boundary will help scale the output value
  7. A feature importance analysis (SHAP etc ) will aid in determining the most important aspects of an action type relative to a given goal.

Contributing Time (programming, brainstorming etc)


I’d be grateful for suggestions of any kind throughout this process. From methodologies to feasability and beyond

Contributing Time

I would be grateful to have anyone involved who has any interest: technical it otherwise. Please send me a message and we can chat.

1 Like

Gracious, that looks terrific for one of my projects since multiple factors are involved, and let’s laugh heartily, I am far from smart enough to use what you are describing. However, in case you’re bored, let me just summarize…based on personal experience preventing depression and mood swings, my vulnerability, I have avoided both since 1974 simply by taking turns, with fellow practitioners, giving silent, accepting, and approving attention for spontaneously arising thoughts and for uninhibited experiencing of emotion which can include crying off stress and emotional pain. Science has proved that the suppression of emotion produces tension, chronic tension produces inflammation of the body, brain, and other health problems, tension triggers emotions and activates behavior stored in the brain from previous stressful moments/experiences, and all of this interferes with our reluctance to face reality, seek and evaluate accurate information, think clearly about it, and interact on behalf of everyone on Earth, nonhumans included. Meditation and mindfulness, prayer, etc., are not enough. Our amazing nervous systems need us to experience emotions, not alone, but when we receive another person’s attention, Peer support groups are adequate for this. But we need volunteers to prove that the silent attention that we can give one another and the freedom to tune in to emotions and cry off stress and suffering as needed, will accomplish for our species all that evolution has provided. THANKS FOR READING if you got this far. It’s okay if you don’t respond. Stay safe, be well. Warmly, Pam - 78 and still grateful to be here with everyone, including you.

I completely agree. Social media was supposed to help with this, but made things worse for most. Marc

Hi, thanks for sharing your reaction, Marc. I hear you, but you know, if we’re not going to give each other this type and quality attention offline, we’re not going to do it online. We have tp provide online space dedicated to it and to programs. Later, I’ll pull you up and learn more abou your focus is. The wonderful thing about QS is, everyone here is trying to accomplish something, one way or the other. Your analysis is excellent. My brain isn’t what it used to be :grin::laughing::joy: yet it is so enjoyable to TRY to think about new ideas, processes strategies. Every brain on Earth, from birth, if safe, thirsts to think.

I swear!! So, I didn’t know I could respond in email, always thot I had to sign in. Nice!!


This is a very interesting line of work.
I spent many years on such analysis and tools making.
Programs Life Navigator, Assistant …
Questionnaire Quest Life.
Objective data about your life, habitat and forecast of the results of your actions.
I will provide all the information if you are interested.
Just keep in mind that such knowledge can be very harmful.
You can lose all illusions.


  1. Technically, this task is quite feasible and government organizations are successfully coping with it.
  2. Everything is limited by the fact that human animals are boundedly rational.
    You can know everything you need, but act irrationally.
    Knowledge is one thing, behavior is quite another. And this knowledge will bring you only negative emotions.

Your plan will probably work on short term goals up to 1 year, maybe medium term.

Good educational game.

To your actions I would add:

  • Met resistance;
  • Random important events.
    The onset of risks.

These factors can nullify all your actions on the way to the goal.

Have you taken this method as a basis?
Observational Skill-based Clinical Assessment tool for Resuscitation (OSCAR): Development and validation

© Worth Consulting 2002

1 Like

This sounds great! Would you be interested in getting a head start on your development work?

Our nonprofit built a web and companion mobile app to help people track goals and goal adherence and derive unbiased insights from their data.

We’re working towards an extension library where outside developers and analysts can make their ideas come to fruition without having to build a comprehensive end-to-end interface, work with authentication/single sign-on, deal with integrations like environmental data, health syncs from Apple/Google/Fitbit, and other external data sources.

We are doing this as a nonprofit to motivate people to focus more on the data than the software development. To learn more about the software, please visit our website to see the platform [LLIF Software Use Cases - Live Learn Innovate Foundation].

Let me know if you’d like to set up a conversation with our technical team about how to develop an extension instead of starting from scratch.

Best of luck with your efforts!

Very cool. How do you avoid the sensitization issues that took Pear down?

What do you mean, specifically? Do you have links to what Pear used to be? I’d love to address the question given more background.

1 Like

Thanks for that link!

LLIF software is not therapy software. It’s a self-driven data aggregator that combines features of many popular apps and allows users to consolidate their data lakes found across multiple data providers. What you do with your data is your choice once you’ve consolidated it. LLIF software doesn’t tell you anything that isn’t based on the math behind your data.

For instance, our trend analysis engines only report correlations based on the data you enter. Examples of trend reports could be “you log ‘headaches’ when the barometric pressure changes by 10%” or “you’ve kept to your goal schedule 95% of the attempts made.”

If there’s a contributor who decides to build an extension like I mentioned in this thread above, that developing contributor would need to manage the terms of use and liabilities themselves. However, all user information remains anonymized in transition between the extension and the data storage. I’m not a technical expert, but for more details about how that can be done I’m happy to connect you with our founder, Jim French. He is much more equipped to answer the technical side of this.

Our platform is meant to hold and protect the data used for the analysis, never changing ownership from the individual who generates/creates that data. Like any nonprofit, if the community does not continually support the presence of our organization, then it will dissolve. The data is then destroyed. It is considered a donor-related asset. It cannot change hands under federal law. We certainly don’t plan on breaking the law either.

If we fail as a nonprofit and cannot sustain community support, the user data is deleted, instead of sold, like what happens to private companies and corporations. If we fail, we plan on making the original code that built the software open-source, instead of letting all the time and effort go to waste. It is the best method of protecting people’s data long-term that is available to us as a US-based NPO.

For further information about who we are, what we do now, and what we plan to do, please check out our ‘master plan’ on our website!

I hope that helps answer your “what if you fail like Pear” question, and if I misinterpreted it please let me know.

Confused. Is this for personal data? I work at a University pain clinic with patient data. Marc

what do you mean by this?

How will you detect dependence between two time series? ML techniques like random forest I presume but how will you make features out of them? Finding relations between variables in time series - Personal Science Wiki

Are these the right indicators for your site?

No, those numbers don’t match our analytics at all. I am curious how they got to such an incorrect representation of visitor analytics. It could be a lack of control for bot crawlers.

Here’s how they claim they gather the data.

It doesn’t seem reliable and has proven inaccurate when researching competitors in my prior roles.

1 Like

Public detailed annual reports of NGOs on their activities significantly increase trust and close many questions from potential participants.
Unfortunately, such NGO reports are rare.
Many organizations are not transparent and often bluff :frowning:

We do our best to provide financial reports and transparency!

You can find our financial reports online here (Financial Reports - Live Learn Innovate Foundation).

We also keep our Candid/Guidestar profile available here (Live Learn Innovate Foundation - GuideStar Profile) fully updated to provide that information to the public.

Further, if you want to find more information from the IRS directly, you can search our EIN on this website directory to find more. Unfortunately, their search does not allow me to link directly to our nonprofit, so you’ll have to enter the EIN (82-5300524) from here (https://apps.irs.gov/app/eos/).

If there’s more you may be seeking, please let me know. I can make sure that information is readily available.

1 Like

Thank you. I have looked at all links.
I did not find the Annual Activity Report on the website.
I just want to understand everything in detail.

Thank you for the input! I’ve since brought up the desire and frankly need for an annual activity report for our nonprofit. I’ll also discuss the topic with our board during our meeting next week.

Hopefully, soon, I can come back to you with a link to it!

Though its certainly less formal, it may be helpful to subscribe to our newsletter where I personally send updates on our progress and activity to subscribers and donors. It’s no obligation, but you may find it helpful.

1 Like

It is very likely that your decision is correct.
Discuss this issue with the board.
Fund financial statements are a legally required formality for foundations.
The activity report is for audience engagement and is very important.

I assume that potential clients - will not read your financial statements.
Most users do not have the qualifications and desire to understand them.

I have no doubts about your professionalism.
You will complete the assigned tasks.
I have been subscribed to your newsletter for a long time and I even read it carefully :slight_smile:

Now I see the fund is making some progress, app downloads have doubled in a few months.

Wish you luck.

1 Like

I’ve indeed noticed the best, or most robust resources are diagnostic etc. My main anticipated diffiulty is

  1. As as n increases in t+n the effect will proliferate and in some cases amplify others diminish. My solution right now is to examine categories of actions and categories of outcomes.

  2. Dataset size: will need to augmented by known be by avior outcome relations, which will alter the distribution of parameter weights but some shift can be recovered

  3. Interest. With it being a personal system, I suspect difficulty finding open source developers.

Any suggestions?


1 Like