How to make AI work in Skill evaluation
Introduction
#ArtificialIntelligence seems to have given me a second wind with #Skill #evaluation, where I was using wood working to get to skills of a #candidate , read about it here, I am now using Music analysis to provide inputs to a #Fuzzy #Logic system.?#Music makes a perfect #companion for Skill evaluation, with the Librosa library of #python providing a lot of modules to work with(I will provide code where applicable).?You can detect the frequency, amplitude, music scale and music notes easily and these form the #input parameters to the Fuzzy Logic system.?
Interestingly, it always requires a real-world skill (like woodworking or guitar music) to evaluate skills, the reason for this is self-evident, the Artificial Intelligence library will have to generically parse the information coming to it. The same Skill evaluator, that handles #EPC skills would be unable to handle #Microsoft #Excel skills unless we had a common real-world skill that would provide #consistent inputs to the #AI system.?Getting ahead of the thought, it is also the way to have AI mimic human intelligence, especially now that Artificial Intelligence (#neuralnetworks) is generating #learning #algorithms independently.?The real-world skill required should have the following traits:
?
1.????The real-world skill should easily be computational through an #AI library, like #guitar music analysis has the #librosa #library.
2.????It should be complicated enough to handle the vast region of skill evaluation - the guitar has the complexity.
3.????It should be logically constructed and should lend itself to break down into respective parts, guitar music has the almighty logic of music theory behind it.
While still at it, I had trouble with wood working as the real-world skill, as it had no ready library to detect cut lengths and ideal structures (dovetails, MITRE joints etc.).?I had used #computer #vision then and it was inadequate for the task at hand.
Further to this, is the thought of Fuzzy Logic or Neural Networks as the AI platform, while Fuzzy Logic has an easier learning curve, Neural Networks has the learning algorithms to back it.?A decision on which platform couldn’t be taken without evaluating both or even a combination of them, the Neuro–Fuzzy system, though it's obvious that the AI system must learn.
?
The Real-world skill
?People often have in the past questioned me about this topic, and the questions are normally as follows:
1.????Why do we need wood working or the guitar to evaluate skills?
3.????Don’t you think you are overdoing the real-world skill part?
I have been known to be circumspect about the answers to these questions, and still am, but I think they need to be answered, nonetheless.?Artificial Intelligence had two courses that it took:
1.????Mimic human intelligence
2.????Use brute force
?
AI initially was thought to be on a route that would mimic #human #intelligence, but as things didn’t work out that ways, AI started to use #bruteforce algorithms, which were data and computationally intensive, very unlike the human #brain.?Everything from there on has been built on brute force, making it less democratised.?#Democratisation of AI would lead to lowering of cost and land the technology in the hands of everybody.
?The real-world skill comes into focus here, it becomes important initially to have a human brain to do AI's work(real-world guitar skills) and then let AI take reigns of the entire product.? A real-world skill takes on the onus of providing:
1.????Consistent inputs to the AI system
2.????Leaves the task of finding what parameters to process to the real-world skill.
3.????We use human intelligence to map out parameters.
?
I have drawn a diagram as below to make the thought even clearer and of course more on this is available here:
The two terms here that seek clarification are inconsistent input and consistent input, it implies that the AI black box to be of relevance in mimicking human intelligence will require consistent input – whether you are evaluating software coding, empathy, excel graphs or even EPC skills, the input pattern should not change. With the Neural Learning algorithms at hand, once we have a few 10,000 samples of individuals to look at, Artificial Intelligence will learn from human beings to handle inconsistent input.
?What are the skills that a Skill Evaluator possesses
A skill evaluator is a person who evaluates skills of an employee or a new hire as also the Artificial Intelligence software product which is used to evaluate skills.?The skill evaluator(person) needs to have this unique domain insight for multiple areas of work.?To evaluate software workers, for instance, would require experience in writing software as well as the real-world skill.?There aren’t many Skill Evaluators, the reasons are these:
1.????The requirement for a real-world skill.
2.????The absolute nonchalance towards Artificial Intelligence.
3.????Artificial intelligence will give us what we need when it can, till then we carry on as usual.
We need skill evaluation because we need products that would jump the AI curve, a real-world skill allows us to jump the curve.?Below are the skills of a Skill evaluator (the person):?
1.????Works through a real-world skill.
2.????Allows convergence.
3.????Knows different domains.
4.????Can evaluate within 10 minutes.
5.????Has the skill of emotional intelligence to know what is going around them.
6.????Can price the product for internal as well as external organisational use.
Skill evaluation vs Experience evaluation
An ongoing debate these days, is about Skills versus Experience, and though I don’t doubt the efficacy of normal HR methods (CV scanning, Telephonic interviews, Face to Face interviews, on-site testing) of evaluating experience, I have had this feeling for some time, that we just aren’t giving the employers the employee-Job Description fit that they want.
Experience is an important part of any employee/candidate’s repertoire, and I don’t want to ignore it.?Though important, I think it is over-rated when it comes to hiring junior and mid-level employees.?The skills that the junior level employees need to be worked on are:
·?????Verbal Reasoning, Comprehension Ability, Logical Reasoning.
·?????Excel/word/powerpoint skills.
领英推荐
·?????Minor Database skills.
·?????Communication skills.
·?????Soft skills.
?
And mid-level (technical or management)employees need to be evaluated for:
?·?????VR, CA, LR.
·?????Business skills.
·?????Supervision and employee handling skills.
·?????Presentation skills.
·?????Analysis and Interpretation skills.
·?????Understanding and empathetic skills.
·?????Experience, as an afterthought.
?
With these thoughts in the back of my mind, let me assure you, that a skill evaluator(AI software product) can evaluate the experience of a mid-level employee or a senior level employee, but it isn’t required till it is required.?It is required when you don't get the Employee-Employer fit and organisations aren't being as productive as they can be. Experience evaluation is a tough job and shouldn’t be left to on-site testing (the NexGen HR product range).
A senior level employee, like a CEO, needs to be evaluated for the following skills:
·?????Core business goals.
·?????Finance and Profit & Loss skills.
·?????Mergers and Acquisitions.
·?????Leadership skills.
·?????Mentoring skills.
·?????Vision/Mission.
·?????Experience.
How does the Skill evaluator evaluate for Experience
The Skill evaluator would evaluate for experience by leaving the route of old treaded paths, by leaving whatever melody I was checking them with on the guitar, my real-world skill.?I would ask them to start working on the experience the senior level or mid-level employee has on the CV, by getting them to talk about their victories and failures and make a new melody while they work and talk, the melody would be rich and new, the uncharted path.?A new melody would exemplify what they are in terms of the following experiences:
?
Once they begin to talk and work, I have ample evidence of what they have attempted all through their lives, professionally and personally.?Evaluating employees using the Skill Evaluator(the AI HR product) leaves no stone unturned, once the new melody is in the bag, the idea is to run it through a proper check, and this would involve (very briefly here), though there is more on the website kawapeople.com:
?
·?????Pitch or Frequency.
·?????Amplitude or loudness.
·?????Structure.
·?????Scale accuracy.
·?????Notes accuracy.
·?????Duration of notes.
?
With these parameters to check for, and feed into the Fuzzy Logic system, it becomes a game of music analysis.?The real-world skill generated parameter-value pair will rank the experience on the CV on a scale of 1-100 through a Fuzzy Logic system as shown below:
?
?In this diagram, as I have been stressing, Experience is evaluated by the real-world skill of Guitar and the parameters generated by the new guitar melody are fed into the Fuzzy Logic Black box.?This black box will deliver 1 single output no matter which skills you want to evaluate, be it Firefighting experience or Creativity experience or People handling experience or even Leadership experience.?This is the advantage of having a real-world skill parameter that won’t change its output parameters and a Fuzzy box that uses Mamdani Inference method to generate a Sat level with the Experience.
?
Conclusion
?With rapid increase in HR data, it is possible that we would make mistakes in the core generalist topics of:
AI would increase the output of managers and employees of the organisation, by cutting short the time to evaluation as also as also the accuracy and precision in all the above 4 points. It therefore, becomes all the more important to be able to give accurate or should I say, truthful data to AI. It just won't do to have AI run berserk on falsehood, it is the opening of the doors to evil. Without sounding melodramatic, we might want to show AI how to handle Skills before letting the Learning algorithms take control.
?
References
1.??????https://aibluedot.com/