Mastodawn

@nitashatiku Elizabeth Holmes and SBF score similarly psychologically. Let’s just say they’re disconnected from any sense of humanity… like - Thanos snap on 1/2 the population and wouldn’t care. VCs have a fiduciary responsibility to make sure these profiles don’t end up lurking in their portfolios. It’s preventable.

Cypher Dec 22, 2022

@KarrieSully @nitashatiku what are you basing your psychoanalysis on?

@cspcypher @nitashatiku trait-based psychology has a deep connection to language. We use ML (based in computational linguistics) on publicly available speech and writing samples. All we need is 50-75 words for high accuracy personality trait assessments.

Simon Lucy

Ah witchy, witchy; who audits your model?

Ambiorickx Dec 22, 2022

@KarrieSully @simon_lucy @nitashatiku @cspcypher the blockchain, duh.

@ambiorickx @simon_lucy @nitashatiku @cspcypher
Lol - no DLT needed except if we decide to go consumer or with any expansion - we may anonymize & secure individual results via DLT. IMHO - I like security as a use case far more than transactions.

@simon_lucy @cspcypher @nitashatiku audit: a) I’m a stickler for making sure we diverse eyes on any models from data to utility. b) we have a couple of trait-based psychologists that periodically audit outcomes c) we check the model against many / diverse human benchmarks that were reviewing for history and current point in time progress

Simon Lucy Dec 22, 2022

You'll realise that cynicism is going to be a common reaction to "All we need is 50-75 words for high accuracy personality trait assessments.", given the use of Myers Briggs and it's erosion of decision making in resource management in many corporations.

@simon_lucy @cspcypher @nitashatiku I anticipate it. I really only chat about the model with folks that I think will get it (humans are just a tiny bit complex + the tech is complex). It’s hard to distill and people need to process through their questions / understanding in different ways. I tend not to blast everyone with a dissertation that would just make people glaze over. Intellectual curiosity and engagement is always better. :)

Simon Lucy Dec 22, 2022

That's all true, but the difference between snake oil and something you can trust is knowing the ingredients, something of the process and what it's really designed for. Furniture polish might clean your teeth but it's not recommended.

So I might ask what's the point? Claims that either individuals are so narrowly and specifically defined or that language is both complex and simple are going to get 'so what' as a response.

@simon_lucy @cspcypher @nitashatiku We’re plotting where people land today in their own personal / psychological evolution, along a continuum via detailed traits. The machine is meant to be radar > diagnosis. Ex: Don’t use it to automate hiring and firing decisions. As a DS pro - you know that machines/ML are great decisions and tasks. Judgement, however is human.
We’re just trying to help the humans improve their judgment. :)

Simon Lucy Dec 22, 2022

Well if the context fits then the machine learning will drive a reasonable solution, sometimes it's only the smallest variations that means the solution misses. When they miss they tend to miss completely.

Of course if the model just perpetuates presumptions and categories and then the correlation with whatever language triggers with too wide a window then is it more than rolling dice on a character generator?

Which is harsh.

KarrieSully Dec 23, 2022

@simon_lucy @cspcypher @nitashatiku totally fair. I may not have been clear. There’s a mix of both the machine and human interpretation at play here. The machine does the heavy lifting around harvesting data, stats, QA, etc. When it comes to translating that into insights that help leaders, teams, and people - it requires human eyes and judgment to go the last mile so we’re constantly checking it against reality.

Simon Lucy Dec 23, 2022