I-Introduction
A flash back to statistical math class may seem really off for a technology blog. In an era when many are trying to integration machine learning into what seems like everything, understanding the practical fundamentals of these errors is critical in knowing when to apply this emerging technology.<\/p>\n\n\n\n
II-What are Type I and Type II errors?
A Type I error is a positive indication when the correct outcome is negative. A Type II error is a negative indication when the correct outcome is positive.<\/p>\n\n\n\n
III-Why do we care?
Why we care is that ultimately computers are used to make decisions based on the inputs given or data is fed back to an end user and that data is then used by that user to make a decision. Lets look at a pragmatic example:<\/p>\n\n\n\n
Why this error is important becomes obvious. A false negative could have significant impact to health, while a false positive means some additional effort but is likely to be discovered on further analysis. In this example biasing a design towards type 1 errors becomes an obvious decision.<\/p>\n\n\n\n
In another example, we may have a research scenario:<\/p>\n\n\n\n
Which is more desirable to mitigate in this case? It is really hard to tell honestly. That said, a type 2 error may be more desirable because it means less products are missed out on, while a type 1 error may be more desirable to keep poor investment in check.<\/p>\n\n\n\n
IV-Conclusion\/What does this have to do with computers? <\/p>\n","protected":false},"excerpt":{"rendered":" I-IntroductionA flash back to statistical math class may seem really off for a technology blog. In an era when many are trying to integration machine learning into what seems like everything, understanding the practical fundamentals of these errors is critical in knowing when to apply this emerging technology. II-What are Type I and Type II […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[6],"tags":[],"class_list":["post-315","post","type-post","status-publish","format-standard","hentry","category-philosophy"],"_links":{"self":[{"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/posts\/315","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=315"}],"version-history":[{"count":3,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/posts\/315\/revisions"}],"predecessor-version":[{"id":371,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=\/wp\/v2\/posts\/315\/revisions\/371"}],"wp:attachment":[{"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=315"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=315"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/underwoodtechnologies.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=315"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
These are common errors in statistical math. Because “AI” models are driven by statistical math, they are prone to these errors. It is notable that undesirable outcomes in medicine<\/a> and policing<\/a> using statistical models have started gaining notoriety in the news.
As technologists, it must be carefully evaluated when to use any technology. Society and culture are largely used to software and computer systems being deterministic, or that overall it gives a specific, measured, repeatable output based on specific inputs barring any sort of program error or bad data. This is incompatible with software based in statistics, and can be extremely dangerous in the hands of people with a mindset that only understands deterministic software.
We must also be aware that the outcomes of what we do with technology directly affects the lives of people whether they be in public, on the operating table, or whatever they may be doing in their daily lives. As we create technology, we must bias errors towards failing safe and we must leave room for human intervention.<\/p>\n\n\n\n