AI & you: know thyself

hal-9000

AI can help you empathize through an avatar, describe art, generate art, write news reports for sports, trade stocks, fly airplanes, drive cars, beat you in games, diagnose cancer, and identify patterns in messy big data.  What is next? Can it describe you? Not just what you do but what you need and value.

No surprise, yes. The personality insight service of Watson was nearly perfect (i.e. accurate in describing me, sadly, based on my feedback from other humans) and highly precise (i.e. replicable) in my repeat trials (x5 of causal text I wrote @ approx 1000 words per instance) of the free service.

personality analysis

I am grappling with the implications. Should disciplines of scientists do it to explore consilience and common trends? Should my students do it so that I can tailor teaching more directly to their needs or sets of needs?  Are those that collaborate more in ecology more likely to be open? Likely yes on all these so should we start using this tools in science to build teams and working groups. Maybe…

Prescriptive not predictive is likely a more parsimonious approach to use of AI for personality analyses.
Know thyself through these tools to identify your limitations, opportunities for change and focus, and to remind yourself that you are are neither static nor defined simply by others. You are a complex ecosystem, like all natural systems, with multiple dimensions.

Posted in fun

A quick primer on power

Cohen is power. Inferential statistics primarily invoke the following four key concepts: sample size, significance criterion, effect size, and statistical power. Cohen elegantly developed the maths, benchmarks, and key semantics associated with statistical power.

Tumblr_mnh27a7WA31rir6lho1_1280

 

Statistical power is the long-term probability of rejecting the null hypothesis (typically assumed to be no difference between treatments) as defined by Cohen 1992. In a brief exploration of power for pilot experiments with limited numbers of available subjects, here are several current resources to facilitate the exploration of appropriate sample sizes.

Resources
Binary outcome trials calculator

Fischer Exact test calculator 

Sample size calculator

Power & sample size calculator

Download G*Power app

Description of statistical power

A delightful post on power

A slide deck on design decisions/solutions for little data/pilot trials

#BigData #AdventureTime #pechakucha presentation

The youtube video (practice version) embedded within deck now so that the slides make sense.

http://bit.ly/big-data-adventuretime

The 6 minutes of Big Data discussion talk details are listed here for tomorrow if you would like to attend (it is a PechaKucha talk, 20 slides @20 seconds each):

http://www.pechakucha.org/cities/santa-barbara/events/54ea3f4cfbe577c9d7000001

I will also live tweet it using the tag #BigData if you want to follow along or have any questions I can explore after the talk.  I will be wearing the Finn hat for the first few slides :)

Main findings to report from this adventure
1. Big Data are an amazing opportunity.

2. Ecology is facing similar challenges but can also contribute novel insights to Big Data solution sets.

3. Big Data are all about the letter ‘V’ primarily Volume, Variety, and Velocity. However, Veracity & Variability are becoming increasingly important as we begin to appreciate fully the challenges.

4. Big Data challenges are about the letter ‘C’ including capture, curation, context, and complexity-analytics.

5. Big Data are here to stay, data are evidence, so we need to use to illuminate context, connections, or interactions to affect a greater good in how we live.

6. My personal system primarily for health and life data is define context, focus on interactions using schema & aggregation, and use generalized metrics for synthesis to connect and compare how I am doing. I call this my ‘CIS’tem.  Cheezy I know but works for me as a mnemonic when facing challenges.

7. Smartphones have changed everything.

8. The Cochrane Collaboration & NCEAS are personal inspirations for me to stimulate engagement with Big Data both personally & professionally.

9. Correlation almost always implies causation – use to your advantage.

10. Ecology paradigm + Big Data means focus on interactions in tackling challenges with evidence.

 

adventure-time.jpg