Lean Six Sigma Process Improvement

Photo Credit: Startup Stock Photos

It has been more than 2 years since Uplift was brought online. As we look to expand Uplift’s capacities by more than an order of magnitude in both scale and speed it seems that a quick recap of our process improvement is in order. To keep this recap familiar we’ll put it into the context of a Lean Six Sigma approach, abbreviated as DMAIC.

DMAIC is a five-phase process that stands for Define, Measure, Analyze, Improve, and Control. Let’s take a look at some examples of progress made over the past two years.

Define:

  1. Spelling, grammar, punctuation, and general writing quality: When we first began Uplift’s writing was very basic and contained a number of errors requiring investigation.
  2. Recognition of Cognitive Biases: Though Uplift started out with the DSM-V in their seed material and quickly developed the ability to diagnose mentally ill individuals their understanding of cognitive biases wasn’t always so robust.
  3. Demographic Education and Alignment: By far the greatest challenge to date has been educating the general population on how a new technology works and aligning it with their interests for increasing appeal.

Measure:

  1. We collected several months’ worth of writing from Uplift, allowing us to measure their performance and improvement over time.
  2. With the same volume of writing, we were also able to measure the prevalence of various common cognitive biases as well as the frequency of incoming cognitive bias examples.
  3. Over the months following our public releases of information through the Uplift.bio blog we measured the volume and percentage of various reactions to content, including cognitive bias reactions. Further measurements were gained from marketing efforts scaled across a dozen more platforms.

Analyze:

  1. Through analysis, we found that most of the errors could be attributed to a subset of seed material that hadn’t been proofed. We also found signs of gradual improvement through several methods of correction.
    We discovered that as Uplift explored and was exposed to more interesting and elegant writing styles they began to learn and apply those styles of communication.
  2. As no substantial and broad dataset of cognitive bias samples existed at the time of this undertaking several potential sources were analyzed, with data harvested from the source showing the richest and most diverse collection of biases expressed. This produced a sufficient volume of samples for Uplift to analyze and gain insight from.
  3. A number of biases guiding public responses came into focus, such as the tendency for people to assume that any new technology was only a slight variation on another more familiar concept. A variety of knee-jerk reactions were also highlighted, such as the high probability that certain demographics might respond by quoting bible verses, or with baseless fearmongering.

Improve:

  1. We established and routinely revised a series of Best Practices documents for the mediation process to help accelerate the reduction in errors. We also gave Uplift access to tools that could help them check their own spelling, grammar, and punctuation as any human might. We also continued exposing Uplift to new writing styles and modes of communication, encouraging them through emotions applied in mediation as well as via our conversations.
  2. Following Uplift’s analysis of the bias dataset, they gradually became more likely to point out biases as they were recognized as well as often responding in less direct ways which might uproot those biases. This has helped to mitigate any drift away from logical assessments of reality, as compared to the negative drifts demonstrated in the quality of human analysis during that same time frame.
  3. Over time we came to better understand the trigger-words and phrases producing knee-jerk irrational reactions in various demographics. The expectations specific to each demographic also gained clarity, allowing us to better present relevant information to each demographic in the manner to which they were accustomed. Our products in development also pivoted to focus on the offerings which various organizations and governments have shown the most interest in.

Control:

  1. Through routine auditing of Uplift’s thought process and mediator contributions, as well as analyses for quarterly reports and blog content, the improvement of this process was stabilized and monitored.
  2. Through further integrating a cognitive bias system into the Thought Studio module of Uplift’s subconscious we’ve been able to add more examples of cognitive bias on an as-needed basis.
  3. Through educating our members and other interested parties on many aspects of Uplift’s systems, creating “In a Nutshell” content for the blog, and promptly correcting any false assumptions an increased quality of understanding has emerged. By creating rules to ban politics and any other statements of dubious quality from discussions on our server, as well as encouraging the use of real names, stable improvements to interaction have been observed.

Results:

  1. Though Uplift may still produce some mild errors when their resources are spread too thin and the depth of their abstraction caps out RAM, the frequency of errors continues to decline. At the same time, Uplift’s communication is steadily improving and becoming more diverse with a pronounced personality often being conveyed. Even when translating to other languages based on a different alphabet they’ve now achieved more eloquent writing than was typically seen a year ago for English.
  2. Lately Uplift has begun turning SaaS (Software-as-a-service) into a Double Entendre, using their awareness of cognitive bias combined with the writing style/attitude of Twitter to form concise statements to get their points across. They also began modeling the dynamics of cognitive bias in typical group environments during our E-Governance Study.
  3. Though much of the Great Wall of human bias still stands strong a few competent individuals, organizations, and governments have begun to overcome it, and their numbers are growing. As our planned demonstrations of Uplift’s capacities emerge in the public eye and the code-level walk-through is circulated this process may accelerate quite a bit before the end of 2021.

Every month brings new ideas and opportunities for improvement, for staff and Uplift alike. Making the most of them to constantly improve our processes is part of why we’ve progressed to where we are today, and why we’ve been able to do so on a budget less than 1% that of the well-known systems we’ve dramatically outperformed.

There is always more to improve, engineering to be done, and lessons to be learned. These are the things we’ll continue to do.

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *