Trending December 2023 # What To Know About College Adjunct Teaching # Suggested January 2024 # Top 17 Popular

You are reading the article What To Know About College Adjunct Teaching updated in December 2023 on the website Moimoishop.com. We hope that the information we have shared is helpful to you. If you find the content interesting and meaningful, please share it with your friends and continue to follow and support us for the latest updates. Suggested January 2024 What To Know About College Adjunct Teaching

Many K–12 teachers are attracted by the idea of teaching college classes in addition to their existing teaching role. There are a variety of reasons why a K–12 teacher may wish to become an adjunct professor, and a variety of ways that doing so could benefit their career.

The hiring of adjuncts, or part-time college instructors only responsible for teaching individual classes, has grown over the last 40 years. More students are going to college, and there is a demand for more instructors. According to an American Federation of Teachers Higher Education Data Center report in 2013, there is a heavy reliance on part-time faculty. K–12 teachers make great adjuncts because of their deep knowledge of topics and fantastic teaching skills. This piece explores the benefits and insider tips to pursuing adjunct teaching.

Adjuncting Basics

Benefits: Why might K–12 teachers choose to serve as an adjunct?

First, it offers unique professional development. Depending on your background, you may want to look in an education department or a department in your content area. Teaching for colleges and universities may give you opportunities for reduced or reimbursed college credits, access to teaching workshops, and the ability to obtain training and research materials via their university library. Some institutions have designed adjunct networking and mentoring systems, but resources differ across universities.

Adjuncts have a major impact on student learning. Students are more likely to take a second course in a discipline and earn a higher grade in the next course when the instructor is an adjunct. Adjuncts are also likely to teach students who need the most support. And students can learn a lot from practitioners—learning about teaching from current classroom teachers is invaluable to students.

Finally, you may have more academic freedom adjuncting than you do in your K–12 position. You will often design the course materials and assignments in the class. You can also choose to accept or decline the teaching contract each semester.

Education and skills: Although there are differences across states, most colleges and universities require 18 graduate credit hours in the subject area in order to teach on the undergraduate level. Some colleges, particularly ones with dual enrollment programs, may provide tuition remission to high school teachers in order to help them reach this 18-credit mark.

However, a graduate or terminal degree will make you more competitive for these positions. I have taught both undergraduate and graduate courses based on my psychology graduate credits and courses in special education based on my postgraduate credits in that area.

Differences from K–12 Instruction

There are also differences between teaching K–12 and teaching college. Here are a few I’ve noticed:

Time: As you probably remember from your own college days, class sections meet for different amounts of time—you may be scheduled to teach a class for three hours one night a week or for an hour three times a week. It’s important to talk with the chair to see if they can arrange classes around your full-time teaching schedule.

Feedback: Students are routinely asked to complete online course evaluations in college. As an adjunct since 2010, I recall only one course observation, but many universities require evaluations in all courses each semester. Your feedback may be solicited as well; some universities conduct yearly staff climate surveys on topics such as curriculum accessibility and campus safety.

Communication: There are communication differences too. Although K–12 instructors share student records with parents, FERPA, a student record privacy law, prohibits sharing records with the parents of adult students.

Adult learners: Finally, there are different expectations when teaching adult learners. The class discussions are rich and often derived from the work experience and life experience of your adult students. You may see a passion and intentionality in your adult learners as they are choosing to participate in higher education.

Communicating about logistics

Usually, the department chair will be your main point of contact as you teach. They help orient you to the campus, teaching schedule, and work expectations. Typically the chair provides information on campus resources and serves as a bridge between the adjunct, the administration, and other faculty.

Scheduling: The chair can help you understand the logistics of the course schedule. They can tell you about what classes they have that might need instructors and whether those meet online, hybrid, or face-to-face. For example, departments like science-based programs, culinary programs, and engineering include labs and hands-on activities that mandate a face-to-face format.

You’ll want to consider your own availability as well, including the number of extra classes you might be interested in teaching.

If you enjoy serving as an adjunct, remember that the contract does not automatically renew. Adjuncts must communicate their continued interest each semester. Typically there is a process in place to make course requests two or three months before the start of the semester.

Format: If you are exploring online classes, it’s important to ask whether the course will meet synchronously or asynchronously. You should also consider your digital literacy in regard to creating and posting videos and using a variety of learning management systems (such as Blackboard and Canvas) or incorporating interactive features (such as breakout rooms and class chats). Universities welcome adjuncts to participate in any IT-related training offered to staff throughout the academic year.

If you are exploring face-to-face instruction, you may want to ask about practical elements such as parking locations, safety (when walking on campus at night), and if you’ll have an on-campus workspace.

Final considerations

There are a variety of resources you can use to ensure that adjunct teaching is for you. Resources such as the American Association of Adjunct Education, the American Association of University Professors, and chúng tôi are useful for new professionals navigating the adjunct role.

Before committing to an adjunct contract, you should also research the institution to determine if it is a good fit. For example, as a Black educator, I explore how a university embraces diversity and inclusion (based on the data on percentages of adjuncts of color). Learning about resources and training that the school provides to support the hiring and retention of educators of color is meaningful to me.

It’s never too late to explore adjunct options at your local college and beyond. I am grateful for the opportunity to adjunct, and I hope to continue to build professional skills and improve the academic experience of my psychology and special education students.

You're reading What To Know About College Adjunct Teaching

What You Need To Know About The Telegram Messenger

Telegram Messenger is an instant messaging app that uses phone numbers (or optionally usernames) to connect users. With Telegram, users can send messages, pictures, videos, audios and any other type of file. The instant messenger app category is a hotly contested field, and some of Telegram’s competitors include WhatsApp, Google’s Hangout, BBM, WeChat, Facebook’s Messenger and LINE (among many others).

About Telegram Messenger

Telegram Messenger launched in 2013, created and backed financially by Russian billionaire brothers Pavel and Nikolai Durov, the founders of Russian social networking site VKontakte who wanted to build a communications system that couldn’t be accessed by the Russian security agencies.

It is built on the open source MTProto protocol. Telegram boasts of being fast, reliable and available on (and accessible from) multiple devices at the same time. It also uses a distributed server architecture with multiple data centres spread out in different jurisdictions around the world.

Telegram is available for Android, iPhone, Windows Phone, Linux, Windows, OSX and via the web.

Features

To begin using Telegram, you must register with a phone number. The phone number then becomes the your identity on Telegram. In addition, you can set a username through which other users can connect and chat without having access to the phone number tied to the username. Usernames are optional and can be deleted at anytime by simply saving an empty username.

In addition to text messages, users can send files of any type up to 1.5 GB. Telegram messages also indicate the status of messages sent. One check signifies the message has been delivered to Telegram cloud, and the recipient has been sent a notification, while two checks indicate the recipient has read the message.

Messages sent with Telegram are encrypted by default, between the user device and telegram servers, which guards against man-in-the-middle attacks. In addition, the messages and files are stored encrypted on Telegram servers, while the decryption keys are stored in Data Centres in other jurisdictions. This way, neither Telegram staff nor government agencies with subpoenas can get to user data.

For the truly paranoid, Telegram has Secret Chats which are conversations between users that make use of end-to-end encryption. That is the chat is visible only to the two parties involved in the discussion, and even Telegram servers cannot view the contents of the chat. When a secret chat is set up between two devices, Telegram helpfully generates a picture visualization of the encryption key for the chat, which both users can compare, and if the image matches, then the chat is secure.

For secret chats, a self destruct timer can also be set on messages. This has to be set before the message is sent. The timer begins when the recipient views the message, and when it runs out the message is deleted from both devices. Photos sent with a timer less than one minute can only be viewed while holding a finger on them, and helpfully Telegram notifies the sender if a screenshot is taken.

Users can create groups containing as many as 200 users and can send broadcast messages to as many as 100 contacts. The difference between groups and broadcasts is that in a broadcast message recipients don’t know about each other, while in a group all participants see all other participants.

Conclusion

Obaro Ogbo

Always on the look out for easier methods of solving complex problems, especially using computers. Obsessed with everything software related (languages, operating systems, frameworks, etc).

Subscribe to our newsletter!

Our latest tutorials delivered straight to your inbox

By signing up, you agree to our Privacy Policy and European users agree to the data transfer policy. We will not share your data and you can unsubscribe at any time.

What To Know About The Latest Cybersecurity Bug In Log4J

On Saturday, the US Cybersecurity and Infrastructure Security Agency issued a statement about a serious new software bug that could impact Apple’s iCloud, Microsoft’s Minecraft, Baidu, IBM, Amazon Web Service, and others. Hackers could potentially exploit this vulnerability to take over websites.

“We have added this vulnerability to our catalog of known exploited vulnerabilities, which compels federal civilian agencies — and signals to non-federal partners — to urgently patch or remediate this vulnerability,” CISA director Jen Easterly wrote in the statement. The bug pertains to something called log4j, and one way that software engineers can protect websites from it is to upgrade to the latest version of log4j (2.15.0). “We are proactively reaching out to entities whose networks may be vulnerable and are leveraging our scanning and intrusion detection tools to help government and industry partners identify exposure to or exploitation of the vulnerability.”

The vulnerability was first discovered by Alibaba’s security team. Here’s what to know about the exploit and log4j.

Log4j is an open-source tool used by Java programs for logging, or creating a record of everything an application has done. (Open-source tools are free and available for anyone to view to highlight bugs or vulnerabilities.)

“You want to create that record for a variety of different purposes, like being able to debug the application if something goes wrong, or be able to understand anything interesting about how the application was used,” explains Shuman Ghosemajumder, the global head of artificial intelligence at F5, an internet infrastructure and security company. “You can create your own mechanism within your own website or mobile app to record that information, or, you can use a logging program created by someone else, [like] log4j.”

When information is passed to log4j, it commonly has to go through the website on which log4j is performing those logging operations.

However—and here’s where this serious bug comes into play—if someone sends the library a command in the form of a special string of characters tucked within that data, instead of just logging that information, log4j will execute it as though it is code in a program.

Think of the string as a skeleton key that opens up the program and allows any attackers to insert their own program, that they control, on that website’s server. In theory, they could run software that allows them to completely take over that website or application.

Additionally, attackers can scan all of the websites on the internet to try and find ones that are responding to this special string of characters.

“This is what’s called a remote code execution attack,” says Ghosemajumder. “One of the things that is particularly dangerous about this is that it can give a cyber attacker a very high level of access to websites and to your accounts.”

For example, hackers can bypass the normal mechanisms that are required to do things on your account, like logging into a bank website or an email account that uses log4j. Because it’s possible that attackers can access private accounts without having the login, Ghosemajumder says that consumers should monitor for unusual activity on accounts that are important.

As for companies and organizations, other than updating the software, they can also use cybersecurity tools to filter traffic going to their website to look for that string and prevent it from reaching log4j. “This is what cybersecurity teams everywhere are doing right now,” says Ghosemajumder. “Hopefully, they’re doing it fast enough for most people to be protected.”

Q&A: What You Need To Know About Google’s Call Tracking Offering

Last Monday Google announced the launch of their basic call tracking solution for AdWords. Since then, my company has received hundreds of questions about Google call tracking from media, clients, prospects, competitors, and random people on Twitter. Google is calling their version of AdWords call tracking ”Website Call Conversions”.

In this post, we’ll answer a few of the questions we’ve received and explain who should use Google’s call tracking, and who shouldn’t.

What did Google Just Release?

Google will now display unique phone numbers on AdWords landing pages dependent on the visitor’s session. In other words, every visitor will see a different phone number on the AdWords landing page. Google is essentially offering session-based call tracking for AdWords only. It is free and, again, it is only for AdWords.

It does not work for Google organic search.

It does not work for the Google Display Network.

It does not work for any lead source anywhere on the web, except AdWords.

From the Google AdWords blog:

What are the Pros of Google’s AdWords Call Tracking Service?

It’s free. Free is free is free.

It works perfectly with AdWords. It is automatically integrated with AdWords and UA.

It is relatively easy to set up…not necessarily easier than a third-party call tracking number, but simple.

Keyword level call tracking for AdWords

Phone number appearance can be formatted to match the website design

What are the Cons of Google’s AdWords Call Tracking Service?

Again, this list of cons comes from agency blogs, industry experts, and the limitations of Google’s AdWords call tracking platform:

It only works for AdWords. Most marketers use AdWords as only a part of their broader marketing mix. If calls result from a Google organic search, Google’s display network or ANY other source, Google call tracking simply won’t track that call.

No call recording.

No in-depth call analytics. This is the big one for our company. Basic call tracking is not powerful. Deep call analytics—conversation analysis—are extremely powerful. That’s where the power of phone calls truly lies.

No local numbers, only toll-free numbers are available. This is a problem for small businesses and enterprises with a local presence.

No telephone features like call routing, scheduled routing, IVR, and geo-routing. These features matter to businesses.

Are Call Tracking Providers in Trouble?

No. Call tracking providers who provide more data than merely basic call tracking are going to be just fine.

Analyzing call conversations is and always has been far more powerful than simply telling you if someone called a phone number or not. That’s rather rudimentary stuff.

Will there be some small prospects or agencies that decide to use Google’s AdWords call tracking? Absolutely. But, will they simply ignore the call data generated by their other marketing efforts? Certainly not.

To quote a prominent marketer I had an email exchange with:

Who Should Use Google’s AdWords Call Tracking?

If I was a small business spending $800/month on AdWords, and that was my only marketing spend, I would recommend using AdWords call tracking. It’s free. My call tracking vendor brethren might disagree with that statement. But, why wouldn’t you use it for that limited amount of data?

Google provides basic call tracking for free for AdWords. If I’m a small business owner I likely don’t need all of the deep data, recordings, IVR, and routing capabilities provided by LogMyCalls and some other call tracking companies. Instead, I just need to know if a call was made.

Small businesses—if they’re marketing exclusively on AdWords—should use Google’s call tracking platform.

Who Shouldn’t Use Google’s AdWords Call Tracking?

In the days since Google’s call tracking release a consensus has started to build, marketers with a small AdWords budget could significantly benefit from Google’s AdWords call tracking. Agencies our company works with—and the blogs of agencies we don’t work with—are saying they will encourage their smaller clients to use Google’s AdWords call tracking. The data is basic, the information is simple, it is relatively easy to implement.

But, for medium-sized clients and enterprises, agencies will encourage them to remain with a third-party call tracking provider. As one agency exec told me last week:

“The data Google’s call tracking provides is just so basic. Sophisticated marketers want more data than that.”

Perhaps Acquisio’s blog says it best:

“Thanks to Google’s call tracking limitations, current call tracking vendors can rest easy. The limited scope of Google’s call tracking solution (Adwords only) means that multi-publisher and multi-channel call tracking will continue to thrive. In fact, call tracking vendors will even work with Google to generate phone numbers, so it seems Google’s release is more of a friendly pairing than an industry disturbing rival for call tracking providers.”

So, specifically, what types of companies should not use Google’s call tracking solution for AdWords?

Agencies: Agencies that want to use Google call tracking for small clients and a third-party call tracking provider for larger clients are going to find themselves in a mess of data, reports, and analytics. Don’t use two platforms when you can only use one. It makes life harder.

SMBs Doing More Than AdWords: Earlier we used the example of a small business spending $800 on AdWords as someone who should use Google’s call tracking tool. That is true – if that small business is ONLY spending money on AdWords.

Why?

Because Google can’t provide call tracking for any of those sources. Their tool ONLY works for AdWords.

Enterprises: Obviously sophisticated marketers at enterprises need more data than Google is providing on AdWords. Every enterprise exec we’ve talked to knows this.

Basically, any company that wants deep analytics or even basic telephony features simply shouldn’t use Google call tracking.

What Does All This Mean for Marketers?

Calls are now mainstream for marketers. If Google cares about something, everybody cares about that something.

Google’s foray into call tracking validates the call analytics world. Google now believes that calls matter. The rest of the marketing world will follow Google’s lead and start caring about calls too. Call data and call intelligence are mission-critical for leading businesses. The exciting thing for the call tracking industry is that now, with the entrance of Google, there is a known entity that cares about phone calls. Google cares about calls!

In short: Google might take a piece of the call tracking pie, but they will also increase the size of the pie dramatically. More pie is always good for everyone.

Deployed Your Machine Learning Model? Here’s What You Need To Know About Post

Overview

What are the next steps after you’ve deployed your machine learning model?

Post-deployment monitoring is a crucial step in any machine learning project

Learn from an experienced machine learning leader about the various aspects of post-model production monitoring

Introduction

So you’ve built your machine learning model. You’ve even taken the next step – often one of the least spoken about – of putting your model into production (or model deployment). Great – you should be all set to impress your end-users and your clients.

But wait – as a data science leader, your role in the project isn’t over yet. The machine learning model you and your team created and deployed now needs to be monitored carefully. There are different ways to perform post-model deployment and we’ll discuss them in this article.

We will first quickly recap what we have covered in the first three articles of this practical machine learning series. Then, we will understand why and how “auto-healing” in machine learning is a red herring and why every professional should be aware of it. And then we will dive into two types of post-production monitoring and understand where and how to use them.

This is the final article of my four article series that focuses on sharing insights on various components involved in successfully implementing a data science project.

Table of Contents

A Quick Recap of this Data Science Thought Leaders Series

Auto-Healing Minus the “Auto”

Proactive Model Monitoring

Reactive Model Monitoring

Address the Root Cause, Not the Symptoms

A Quick Recap of this Data Science Leaders Series

In this series on Practical Machine Learning for Leaders, we have so far discussed:

Once the optimal end-to-end system is deployed, do we declare victory and move on? No! Not yet, at least.

In this fourth (and final) article in this series, we will discuss the various post-production monitoring and maintenance-related aspects that the data science delivery leader needs to plan for once the Machine Learning (ML)-powered end product is deployed. The adage “Getting to the top is difficult, staying there is even harder” is most applicable in such situations.

Auto-Healing Minus the “Auto”

There is a popular and dangerously incorrect myth about machine learning models that they auto-heal.

In particular, the expectation is that a machine learning model will continuously and automatically identify where it makes mistakes, find optimal ways to rectify those mistakes, and incorporate those changes in the system, all with almost no human intervention.

The reality is that such ‘auto-heal’ is at best a far-fetched dream.

Only a handful of machine learning techniques today are capable of learning from their mistakes as they try to complete a task. These techniques typically fall under the umbrella of Reinforcement Learning (RL). Even in the RL paradigm, several of the model parameters are carefully hand-tuned by a human expert and updated periodically.

And even if we assume that we have plenty of such products deployed in real-life situations, the existing data architectures (read ‘data silos’) within the organizations have to be completely overhauled for the data to seamlessly flow from the customer-facing compute environment to the compute environment that is used for building the machine learning models.

So, it is safe to say that in today’s world, the “auto” in auto-healing is almost non-existent for all practical purposes.

Let us now see why machine learning systems need healing in the first place. There are several aspects of the data ecosystem that can have a significantly negative impact on the performance of the system. I have listed some of these below.

In-Domain but Unseen Data

A typical machine learning model is trained on about 10% of the possible universe of data. This is either because of the scarcity of appropriately labeled data or because of the computational constraints of training on massive amounts of data.

The choice of the machine learning model and the training strategies should provide generalizability on the remaining 90% of the data. But there will still be data samples within this pool where the model output is incorrect or less-than-optimal.

The Changing Nature of Input Data

In all real-world deployments of machine learning solutions, there will be a subset of the input data which comes from a system that the data science team has little control over. When those systems change the input, the data science teams are not always kept in the loop (happens largely due to the inherent complexities in the data world).

Simple changes in the input data, like a type change from ‘scalar’ to ‘list’, can be relatively easily detected through basic sanity checks. But there are a variety of changes which are difficult to catch, have a substantially detrimental impact on the output of the machine learning system and unfortunately are not uncommon.

Consider, for example, a system deployed to automatically control the air conditioning of a server room. The machine learning system would obviously take the ambient temperature as one of the inputs.

It is fair to assume that the temperature sensors are controlled by a different ecosystem which may decide to change the unit of temperature from Celsius to Fahrenheit without necessarily informing the machine learning system owner. This change in input will have a significant impact on the performance of the system with absolutely no run-time exception thrown.

As the systems get complex, it is almost impossible to anticipate all such likely changes beforehand to encode exhaustive exception handling.

Changes in Data Interpretation

The landscape of just about every business is changing quite rapidly. Words like Dunzo, Doordash, and Zelle, which didn’t exist a few years ago (and would hence be just marked as ‘out-of-vocabulary’), have now become keywords with significant interpretations.

Uber, which used to be associated only with transportation, can now be interpreted as food-related also. Wholefoods, which had nothing to do with Amazon just a few years ago, can now influence Amazon’s financial reporting.

Further along, food delivery, which is today probably associated predominantly with a bachelor-like lifestyle in India, may get associated with working-young-parents-lifestyle in the near future.

What these examples show is that as new business models emerge, existing businesses venture into adjacent spaces, mergers, and acquisitions happen, and the human interpretation of a particular activity may change over time. This dynamic nature of data and its interpretation has serious implications for our machine learning model.

Machine Learning Systems Deployed in Unexpected Contexts

One human capability that is vastly superior to today’s machines is to weave in seemingly disparate sources of information to form a complete context to interpret a data point.

Consider this example from the fin-tech industry:

If we know that a financial account is of a UK resident, then it is relatively easy for both the machine and the human expert to interpret the word “BP” to mean “Bill Payment”. But if the same account holder travels to India and has a financial transaction description that has the word “BP”, human experts can very easily infer from all the context available to them that BP here likely stands for “Bharat Petroleum”.

A machine may find it near impossible to do such context-based switching. And yet, this is not a corner case. As machine learning systems become more and more mainstream, they will be expected to mimic the context-aware human behavior.

While we continue to build systematic ways in which context can be codified into the machine learning systems, we need to build (semi-)automatic techniques to monitor trends in the input and output data.

Proactive Model Monitoring

If we cannot auto-heal, what can be done then? The next best thing to do is to continuously track the health of the machine learning model against a set of key indicators and generate specific event-based alerts.

The obvious follow-up questions are what are these key indicators and which events trigger an alert. These questions are addressed by the proactive model monitoring framework.

The key element of the monitoring framework is to identify which input samples deviate significantly from the patterns seen in the training data and then have those samples closely examined by a human expert.

Unfortunately, there is no universal way of identifying which patterns are most relevant. Patterns of interest largely depend on the domain of the data, the nature of the business problem, and the machine learning model being used.

For example, in the Natural Language Processing (NLP) domain, some of the simple patterns could be:

Identify all the data samples which have at least one word not seen in the training data

Identify all the data samples which are at least N-words longer or M words shorter than the average number of words in the training data

We can quantify the word distribution in the training data using modeling techniques like Gaussian Mixture Models (GMM). Now, given a test data sample, find the probability of the sample given the GMM. All data samples with a probability lower than a certain threshold can be marked as ‘non-representative’ (i.e., anomalous) and sent to the domain experts for further investigation.

Source: ResearchGate

Even more sophisticated patterns for identifying test samples of interest can be devised based on the knowledge of the business problem, the specifics of the data, or the specifics of the machine learning machinery used.

For instance, any machine learning solution can be thought of as a combination of multiple elemental ML components. As an example, a machine learning model for intent mining in a conversational agent may consist of three ML modules:

A module for audio analysis of the raw speech input to identify the sentence type (i.e., statement, question, exclamation or command)

A module for text analysis of the transcribed speech input to identify the semantic message, and

A module that combines the output of the other two modules to identify the intent

During the training phase, we can identify the relative proportion of the paths traversed by different training samples through these three modules and the corresponding predicted outputs.

During the model monitoring phase, we can identify the samples that led to a particular output but the path traversed through the three modules wasn’t one of the paths observed during the training phase for that output.

Note that to achieve this level of pattern-based model monitoring, the end-to-end solution needs to have a robust logging mechanism.

Reactive Model Monitoring

After the successful deployment of a machine learning-driven solution, the data science team will almost always feel like they have earned bragging rights like “our system has state-of-the-art 99% accuracy!”.

But instinctively (and rightfully so), the first thing that the customer-facing teams will ask is “what is the plan to address customer escalations on the 1%?”.

This calls for reactive model monitoring which performs root-cause-analysis (RCA) of the customer escalations and provides an estimate of when the bugs will be fixed.

Reactive model monitoring is quite similar to that of proactive model monitoring. But there are subtle differences in terms of the end goals.

Whereas proactive model maintenance identified general patterns in the test data which are outliers compared to those in the training data, the goal of Reactive Model Maintenance is to identify what led to an erroneous output in a specific test sample and how it can be rectified.

The data science team thus needs to be cautious when accepting the rectifications suggested by the reactive model maintenance process as those recommendations can possibly be detrimental to a wide range of data samples.

Some of the other challenging aspects of reactive model maintenance are that some bugs can be resolved by a simple change in one of the config files while some may need elaborate retraining of the ML model. Also, some bugs may be within the tolerance threshold of a typical user while some maybe what I call as the ‘publicity-hungry’ bugs.

A ‘publicity hungry’ bug is any incorrect behavior of the machine learning system which is totally unexpected from a human expert.

For instance, in an ML-powered conversational agent, in response to the user’s query of “I am tired”, if the agent responds with “Hello Mr. Tired, how are you?”, then that is sure to get a lot of tweets and retweets and similar publicity! Such publicity-hungry bugs need immediate resolution.

The Service Level Agreements (SLAs) will thus need to be carefully crafted keeping in mind the severity of the bug on one hand and the systemic changes needed on the other hand.

Address the Root Cause, Not the Symptoms

Given these wide varieties of sources that may lead to a drop in performance of the ML-systems over time and the intense pressure to fix the issues within a given SLA, it can be tempting to have a ‘thin-layer-of-rules’ which bypasses the ML machinery completely to address the immediate customer escalation.

Such a thin-layer or hot-fix approach is actually a ‘lazy-fix’ which has the potential to turn disastrous in the long run. Thus, such a thin-layer of rules should be touched only under extreme conditions and should not be allowed to get beyond a certain ‘thickness’.

When the pre-defined ‘thickness’ is reached, our machine learning model has to be retrained to address the issues encoded in the thin-layer.

To borrow an analogy from the medical domain: addressing symptoms may not need an expert but if that is routinely substituted for a thorough diagnosis, the situation can precipitate quite rapidly.

Just like accurate medical diagnosis comes from analysis of the patient’s history, proactive model maintenance has to be broad enough to quickly help identify the root cause of a customer escalation.

Retraining a machine learning model that is already deployed in a live production environment is much easier said than done. For one, there are multiple ways to solve a particular data-driven problem, and as we see more data our choice of the model may change.

Secondly, the data science team that built the original model and the team that is maintaining the model may not readily agree on the best way to retrain the model. Moreover, the team that built the original model may have tried out a wide variety of training strategies/modeling techniques before settling on one.

This information is typically not documented and hence model retraining may very well lead to a net drop in the accuracy.

To add to the mix, a lot of the times the end client may prefer receiving consistent output over a now-correct-but-earlier-incorrect output. Here is what I mean by this:

Say your original speech recognition system would confuse “Tim” with “Jim” about 80% of the time. The end client estimated this frequency of error and has included mechanisms in their downstream processing to try both ‘Tim’ and ‘Jim’ with an 80-20 proportion.

Suddenly, when the retrained speech recognition system reduces the Tim/Jim confusion to only 10%, the end customer may not readily agree to make the necessary (potentially non-trivial) changes on their end. The business teams and the customer-facing teams may, in such cases, make a decision that certain customers will continue to get the old speech recognition system while the other customers will be migrated to the newer one.

This means the data science teams will now have to maintain two models! This opens up a whole new area of discussion called ‘technical debt of machine learning models’. Consistency can trump accuracy.

Turns out “Be Consistent!” is just as great a motivating phrase for ML models as it is for humans! An area I would love to discuss more, but not in this series.

What’s in a Name!

“What’s in a name?” – William Shakespeare

Finally, the general perception is that the phrases ‘model maintenance’ and ‘model monitoring’ sound ‘uncool’ compared to ‘model building’.

In contrast, what I have seen is that the level of data science maturity, depth of big data engineering, and business understanding needed in ‘model maintenance’ is the order of magnitude more than what is needed in ‘model building’.

I am always tempted to rebrand ‘model maintenance’ as ‘model nurturing‘ particularly so in the light of the critical role maintenance and monitoring play in ensuring customer delight.

End Notes

If you are in the tech industry, there is no escaping the buzz around Artificial Intelligence, Machine Learning, Data Science and related keywords. I genuinely believe that all this focus on data-driven technologies will help bring in substantial efficiency in existing processes and help conquer new tech frontiers which have long been elusive.

However, the general expectations from these technologies are dangerously unrealistic, largely fed by the popular imagination of sci-fi literature. Part of it is also affirmed by what we see in some of the low-stakes consumer-AI applications.

Data science cannot generate impact in isolation and that the entire organization has to be trained into a ‘data-culture’, which of course is easier said than done, and

Years of concerted efforts by data experts have gone into building the consumer-AI applications that are gaining popularity in the media of late. This mismatch between the expectations and the actual reality is driving us closer to what is termed as the ‘AI Winter’

I am certain that data-driven technologies are the best solution to solve most of the problems that the tech world faces today. But, in the same breath, for these technologies to succeed, we need a holistic approach with the right expectations.

Through this four-article series, I am hoping to share my learnings of bridging the gap between a ‘prototype of a data-driven solution’ and an actual ‘data-driven solution deployed in the real-world with stringent SLAs’. I hope you will find these learnings valuable as you continue your journey on data-driven-transformation.

What You Need To Know About The Telegram Messenger

What To Know About The Latest Cybersecurity Bug In Log4J

Q&A: What You Need To Know About Google’s Call Tracking Offering

Deployed Your Machine Learning Model? Here’s What You Need To Know About Post

What Is Allblk? Here’S Everything You Need To Know About The Streaming Service