Thwarting facial recognition

www.nytimes.com/2020/08/03/technology/fawkes-tool-protects-photos-from-facial-recognition.html

Ethical AI Frameworks vary

EY global study finds disconnect on ethical artificial intelligence (AI) priorities between public and private sector https://finance.yahoo.com/news/ey-global-study-finds-disconnect-230200076.html

Part 2: Ethics in AI

In attempting to come to grips with where ethics is being applied to AI I have found through formal research approaches that both private and public sector organizations and companies have created frameworks or operating principles for ethical use of data and algorithms.

That is the good news.  The bad news is they are not being enforced or really being used.  It appears as though lip service is being paid to the recommendations.

All of this got me to start to think of fundamentals of ethics and how and why it can be applied for AI or cognitive enterprises and that it would work.  

One of the challenges is if you look into the field of Ethics you are delving into a subject that has been discussed, argued and loosely defined over many years.  So ethical theory is not new yet we are attempting to apply it to an area that did not exist.

But that should not be a concern for us. If you look at how ethics as be applied and re-applied to all aspects of humanity over the years.  And in recent decades areas such as equality, abortion, class differences, hiring practices etc.  So AI will be nothing new to ethical theory.

The challenge that is raised is defining what really are ethical principles we can apply that make the most sense to ensure people are protected.

One of the guiding principles I was trained in was the area of equal consideration of interests.  So I will delve into what that means in greater detail in AI in the next post.

Have a great day !

Part 1: A Journey into Practical Ethics and AI

Hello World,

 

I have not posted in a while since I have been a little pre-occupied doing my job.  I read daily and keep so many articles on AI and I see the tide come in and go out on opinion on where AI is in being used and how advanced or how simplistic it currently is.

But one thing seems to be ringing true and that ethical practises in AI need to be addressed.  I here about some companies calling for regulation or “laser” regulation which sounds to be a bit concerning.  

When I work with clients the discussion of data and the ethical use are discussed because without clean and ethically clean data the AI algorithm or digital platform will fail.  Once you get into the AI side of the discussion then you are into areas of bias, variance and transparency.  

So from my perspective I see ethical practices to being fundamental in all aspects of implementing AI.  I have said this many times but regardless of industry or technology you can reduce the discussion to a strategy around People, Process and Policy.

So with the rubric in hand of those three principles I am going to set off on a journey step by step to look at how practical ethics can be applied to people, process and policy in the realm of using AI technology. 

Wish me luck and may the trade winds be at my back.

BTW the images may be a little random but they are mine and they just give you something nice to look at while thinking about applying ethical practices to your data and AI journey.

AI Use Case Can Robotics Help a Law Firm

Apologies for the delay in posting.  Had a house move and busy work schedule that took precedence.

So continuing on examples of how AI would aid this international law firm we now look at how Robotics ( The third arm of AI ) could assist the company to differentiate in the market place.

Organization Process Enhancement with Robotics

Given that the use case being described is for a law firm; this will neede some creativity. A caveat, that I have learned over the years; is that lawyers; do not like technology getting in-between them and their clients.

Lawyer training in communication and arguing. Using a AI technology such as https://www.oraiapp.com/. Integrating this technology into a robot could aid in training. The robot could talk and listen to how they speak. This would to help lawyers imporve thier public speaking. This would also support a strategic imperative. It would reinforce the strength of the industry focused team that they have across the country.

Legal document and case file management. Lawyers have to be exact on how a case file or docket is organized and managed. Since physical documents are still required in legal cases; a robotic system could be used to manage these. The robot could organize and move case dockets and documents throughout the office.

Robotic Process Automation (RPA) for collection, routing and processing of student resumes and supporting documentation. RPA would automate documents coming into the organization. As well as automatic scheduling of first and second interviews of the students. This would reduce time and cost; as well as standardize the hiring co-ordination process.

Cost leadership, differentiation, or focus to give them a competitive edge.

This national law firm could realize specific advantages with robots. An anthropomorphic robot that is able to train lawyers to speak more effectively would reduce cost. Also enhance differentiation and improve focus as market strategies. Automating the training process would be done with reduced internal costs. As well as direct feedback for each lawyer versus classroom training. Each team would be better trained and so have greater advantage in focus and differentiation.

Robotic processes can also function in the background of the firm. File management throughout the office would give administrative staff time to focus on more complex cases tasks. Also, by automating the mundane tasks of resume collection. Coordination and ingestion into the NLP system; described last post, would reduce cost. As well as be a competitive differentiation against other law firms. Intelligent robotic back office functions ensures standardization. Audibility of hiring processes would improve trust in AI. Automation of legal case file management allows the firm to differentiate on market focus and differentiation.

Technical and Managerial requirements

There would be greater challenge with implementing robots. Greater disruption to processes and work culture could occur. Executive participation and oversight again would Manyred. Many surveys or assessments would need to be taken over the roll-out to ensure adoption and use is successful.

For the communication skills training robots a change management team of HR, Practice Leaders; as well as robot handlers would be required. Given that; the AI technology would need to be integrated to a robotic interface. An assessment of the “friendliest” robot would help with how lawyers will interact with it. As for the robot case file manager. IT and Records Managers as well as administrative resources would be needed to train the robots on the entire process of case file management. As well a training plan for the physical introduction to case file moving robots and how they will interact with office staff.
Use of Robots would be outside of the original business and IT strategy. The other two initiatives (Robotic Communication training and RPA for hiring process) would need a change in the IT strategy but not the business strategy since they are both back office functions.

An assessment would have to be made of the communication skills of each lawyer before using the robot. The AI software can segment and record each lawyer so reports and performance improvements can be measured. A tech resource would need to be available for technical issues; since robots need to handled when dealing with human interactions.

We interrupt normal broadcasting for a thought bubble !

Good Morning.

 

I got into my car this morning as usual and at some point I pull up Spotify to play one of my own playlists.  This morning I was a victim of their own marketing initiatives.  Instead of my recent heavy rotation playlists being at the top there was this podcast link to Startup.  I read the blurb and it seemed interesting since I did have a stint as an entrepreneur in starting up a small retail/wholesale business when I finished university.  So I started the podcast and interestingly it was on the live journey of Gimlet.  This episode dealt with the run up to Spotify acquiring them.  One area of discussion that they mentioned was the cultural or managerial differences in running the business.  Matt was focused on strategy and running the company by logic and numbers.  Whilst, Alex was running the creative and content.  One comment stuck out that the word “feeling” came up.  In that if in business review meeting something was discussed that touched on the creative side then the word “feeling” kept coming up.  “That doesn’t feel right”  or that “feels like a good idea” whereas the numbers (sponsorship, subscriptions, revenue and growth) would be counter to what felt good

Like all good marriages there will be this counter balance.  But it got me to thinking that AI projects fundamentally need to align to the strategy of an organization: ( Cost/Value, Industry/Market differentiation etc). However if culture change is not incorporated into the project then we will hit a larger wall.  Organizational change management is always given lip service but when you deal with peoples “feelings” in business there has to be a means to accommodate and or incorporate that into to plan for AI to succeed.

I always reflect back to the fact that tip jars or charity jars in shops always do better if you put ‘googlly eyes’ on the jar.  The put a human face—or at least eyes—on something seems to help in interaction and success.  Robotics is the third arm of AI and we are seeing that if a human interaction element deals with perception, interaction and emotion it is more successful as well.

Bottom line of this thought bubble: plan for and incorporate the broad array of perceptions and emotions into the AI solution design and business case.

 

Now back to normal programming.

AI Use Case 1: Step 2 Future State of HR Hiring

 

So to summarize from the last post.  The first primary step in designing for AI projects is to align to the business strategy. As discussed earlier a specific use case is needed. As well as understanding current state processes, capabilities and outcomes.  Now that we understand the state of the organization and the intended high level outcome. We get into to aligning AI capabilities that could best address the desired outcome.

A US National law firm in leveraging prediction ML models could apply it to these processes:

The hiring process. By improving candidate identification which would then reduce hiring costs and time. HR knows all the variables of good candidates; but cannot do it by hand today in the time needed. Having an ML Model improve candidate identification would help the HR team.

Predicting attrition factors of a hired partner would be another process.  Early intervention would reduce risk and cost associated with losing a law partner.   The ML model could analyze more data points for each partner faster than HR.

Skills management would be another process to automate. Identifying skill gaps by combining performance reviews and other sources of data would benefit each lawyer’s assessment.   Since more data points would provide better insight into where a lawyer would need training. Which would support the market strategy of industry practices and expertise.

Improved HR performance reviews would be another area that would enjoy ML models. Analyzing more factors of a lawyer’s activity; versus semi-annual reviews, would give better insight. Documentaton: time, court time, legal preparation time as well as HR data provide insight into a lawyers performance.

Setting a vision or desired future

The cost to hire, train and keep a skilled lawyer can be high. The efficincies gained by greater insight into lawyer HR data will reduce overhead costs. ML models would predict candidates potential of attrition and skill level matching in a more automated fashion.

Machine Learning will only improve HR assessment and feedback. As well as; identifying early intervention on job dissatisfaction. The law firm would be able to improve it’s work place ratings on a national basis. Thereby; making a more desirable location to work and providing differentiation in the market place (another Porter strategy dynamic).

When the firm is able to hire the best resources at a lower cost. The market focus that they provide will only improve and provide clear competitive differentiation. Which is in alignment to their market strategy of industry focused teams. 

Current state is manual based. And on based on human experience only.  With the ability to increase data points and measurement the combination of ML and HR resources would create an increased collective intelligence. This is in line with the three market factors of Porter: Cost, Differentiation and Focus for the company.

Technical and leadership and managerial requirements

In order for Machine Learning to provide the most benefit requires first. Subject matter experts to assist in the development of the AI project. Thus, Practice Leads interviewing staff, HR leads, existing BI, and tech resources need to take part.   All these roles will need to engage with the development of the model and the testing and training of the model.  There has to be executive sponsorship in order for this to succeed. This sponsorship’s role is to communicate and ensure the project stays on track and schedule but also be seen as involved. 

Technical requirements would ensure that the sources of data such as: HR and Case Management data could be accessed and offloaded to “parameterize” the data sets. The “parameterizing” of the Machine Learning model is critical the success of AI.   Natural Language Processing will extract data from CV and other free-form text sources of information.

Next instalment will look more closely at the NLP aspect of this use case. As well we will look at the need of a “domain” or corpus of knowledge that is needed for the AI to be of any value.

Additional technical requirement will be.  What does the organization have currently.  Does the tech staff have the appropriate training for; BI tools, Predictive Analytics, Data Integration tools.  Another consideration are; can this be done in the cloud or does it need to be done on premise due the sensitive nature of data sets.  What are the security and privacy capabilities of the ML technology? Has an ethical review been done on types of information and sourcing to minimize risk ?  These questions need to be addressed by all of the team members.

AI Use Case 1: HR Hiring and Staff Attrition Step 1 Knowing Current State

In this new set of series I will post and reflect on engagements that I have had with organizations around the world.  I will discuss what they are attempting to do with various AI related technologies and then provide updates on the projects as they progress.  Like all AI projects they tend to start smaller in scale and more importantly with a specific use case or business outcome needed.  For reasons of confidentiality I will not name or describe the actual organizations in any detail.  What is important is to learn from their approach and what succeeds.

Step 1: Understanding their current state

The organization is a large law firm in the US.  They are currently attempting to improve their hiring process to ensure they can hire the most talented people for the various law practices.  Currently they use standard HRMS technology to track skills assessment and performance metrics; but only once a person is hired into the firm.  They currently use descriptive analytics to run standardized reports on historical data only.   They now want to improve how they hire lawyers.

To hire the best talent is a highly competitive process.   Each year the number of law students is limited, and the process is controlled by the law societies.  Therefore, this firm wants to be better at quickly identifying the top talent so that they can interview and potentially offer them a summer placement before their competition does.  The current process is extremely manual in nature and therefore they wish to automate the process which can help reduce the costs of hiring.

Strategic alignment of the use case to the organizations corporate strategy is key in ensuring successful deployment of the AI technology.  The external strategy of the organization is to differentiate themselves with their industry experience and team focus.  Internally the strategy is to optimize cost in the recruitment and retention process. The investment that the law firm makes in hiring and training a lawyer can be high; therefore, they wish to maximize their investment in a person that is hired and employed with them.

Current state is a highly manual process.  The hiring team has to read thousands of resumes and supporting material and summarize and review each candidate.  Then; setting up interviews of the top candidates with the view of hiring a certain number of summer students each year.

On the initial assessment of current state it was felt that targeting the internal strategy of lower cost would be the best area to focus on.  This strategy of lowering cost to gain the best legal resources is where AI can assist. 

The HR and hiring teams have a clear understanding who makes an ideal candidate and having an AI system that can read resumes and other sources of information on each candidate would speed up the candidate identification process and reduce the risk and cost of attrition over time.  This then supports their external strategy of having the best industry focused legal teams for the clients they serve.

Certain questions and concerns around data use and data privacy will need to be discussed as well as ethical considerations.  The HR team that participated in this workshop highlighted the fact that they have a pretty good ‘sense’ on how to identify strong candidates.   However; they were also hoping that as the AI technology learns from assessing candidates that the HR team would learn of candidate dynamics that they may not have considered in the past.  As an example a student applying from overseas may not have the standard experience and educational background that typically is reviewed.  

One other observation in running this workshop which will be echoed across all of the use cases I describe is that having line of business as well as IT in the room at the same time is extremely beneficial.

Next posting will be on future state and where AI can assist.

AI and Ethical use of data: Data Validity Part 1

As mentioned yesterday I had the pleasure of taking a course on Ethics and Data Science.  Given that data science is a key area of the Machine Learning area of AI I thought I would expand on the subject as a starting point.  Each bullet I discuss has a more detailed discussion required.  But I truly believe that Ethical/Business Conduct requirements will be needed for all AI projects in order to provide transparency and explainability.

IMG 1823

So what are the risk considerations for ethics in AI ?  You will notice that there is overlapping considerations that have to be thought of and included.  

  • Data Validity
  • Algorithim Fairness
  • Informed Consent
  • Model Errors
  • Societal Impact
  • Ossification/Rigidity of ML models
  • Surveillance Impacts
  • Managing Change
  • Regression
  • Bias/Variance

So as an example of data validity:  and I am starting to see this criteria being included since it has a high risk or legal ramification.  In this day and age of access to third party data sets ( legally and illegally ) or the data sets that you have collected as an organization.  Are you questioning where the data comes from and if proper “informed consent” was given by individual providing that data or information ?  Did third party organizations thoroughly vet and validate the information ?  Has it been modified or redacted or scrambled ?  Can you still identify individuals or information by extrapolation ?  What proof will stand up in court if you are sued for accessing information not properly vetted by a third party.  Are you moving data from one line of business ( sales to marketing ) and are you violating any agreements that you have with clients or leads ?

 The course I took was done a few years ago and had a very optimistic tone to it that regulation would slow down innovation or drown skills; the recommending direction from the professor at that time was: don’t surprise people with outcomes and be able to explain how the model got to that outcome but leave how and what we analyze to the data scienctist.  I think we will see regulations step in.  Any time you have a practice: medicine, legal, engineering, real estate or accounting regulation has to be in place to protect human rights and the individual human.  

AI and Ethics

IMG 1954

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

On a recent trip to New Orleans for work, we had the privilege of touring the location that houses and builds the floats for the Mardi Gras parade.  I took quite a few photographs but this grouping of money bags and Elvis got me thinking about ethics and AI.

You will be hearing more and more of this issue.  There are many who do not understand AI ( ML, NLP and RPA ) and are therefore concerned about what it will and will not do.

I had the benefit of taking a course on Ethics in Data Science as well as taking a course through MIT Sloan on Business strategy and AI ( I’m half way through so wish me luck ).

There are many factors for success or failure.  I think one key factor is the ability to bridge the gap between technology and business units.  

“Explainablity” is something that you will hear a lot about – the ability to clearly explain what AI is doing and how it came to its outcome.  That but also to clearly show there is no ossification of the system in that a hiring system is not hard wired the hiring practices of the the organization.  Or that young people are being denied loans at a  banking system due the the fact that the model is biased towards older applicants and that it was never given data sets for a population below the age of 35.

Ethics and AI will become more important to the point that we are now starting to talk about regulation and compliance to ensure good use of AI vs uncontrolled use of information and models.

The coin is in the air and flipping I wonder which way it will land ?

Maybe I should develop an AI Maturity Model assessment toolkit, I did this for Case Management systems and Cognitive Computing so I should be able to do it AI Maturity.  And in doing so Ethics and Transparency will have to play a key role.

 

 

 

 

 

 

 

 

 

%d bloggers like this: