The 4 Most Difficult Personality Types for Data Scientists to Work With
Working as a Data Scientist is definitely a privilege, and a role that I have enjoyed for years. While most interactions are professional and enjoyable, there are a few personas that I have experience working with as a data scientist that I think others can relate to. I’ll share what I feel are the 4 most difficult to work with, and give some tips on what you can do to make your workplace more conducive to data science.
The Task Master
When deciding on which color to paint the walls in your kitchen, has anyone ever suggested to paint the kitchen every color and take a picture, just to see what it would look like? A task master would. I mean, if you didn’t try the color “rust-orange-6”, how do you really know it wouldn’t look good in your kitchen?
A task master won’t offer direct steps to complete in support of your goals, but they will offer permutations or combinations of tasks to churn through. The line between the list of possible tasks and the end-goals is thin…perhaps dotted. Brainstorming and throwing ideas out there is fine, but stating that all suggestions are requirements and that this is sufficient as leadership or guidance; that’s what a task master does.
Try to ask how completion of all the tasks will improve the situation, and try to present a clear relationship between time, scope and quality. (Larger scope in less time means low quality). Offer a completion of a subset of tasks as a token of goodwill before dismissing the laundry list of tasks.
The Hand Waiver
Nobody is expecting a connect-the-dots approach to data science projects. Part of being a data scientist is knowing or discovering the method by which to get something done, and there is sometimes ambiguity in how a current task will directly benefit the project. Difficulty arrives when large, sweeping, and usually vague assignments are given to the data science team as direction, and the person delivering this assignment either won’t or can’t offer details.
The hand waiver is someone that is going to give the team vague ideas without detail, and they are going to respond to requests for detail with deflections. For example, I could say “you should fit a model that improves our business outcome using the best features relevant to the client”, but what if someone wants to know how to find the features relevant to a client? Well, if my response is “just use an appropriate feature importance technique, it should be easy to find online and it shouldn’t take long”, then no more detail has been given and I’m not trying to answer the question. This makes working with a hand waiver difficult, since follow-up questions for details are made to seem like a lack of understanding or competence.
Try to rewrite vague requests in easy to understand terms that people understand and show the value of doing so. Your hand waiver colleague will may take notes. Don’t waste your time trying to track down vague instructions, but focus on well defined tasks instead and make it apparent. Don’t start hand waiving as a reflex, be more explicit in how you bring ideas to the table and show others how you want ideas to be communicated. Also, don’t let this kind of behavior get you down: you have the skill set needed to solve problems.
The Distortion Artist
These artists of disinformation can either be coworkers or recipients of your work. They will take your masterpiece and go all Picasso on it. Either by mistake or willfully, a distortion artist wants your data science work to say a certain thing and try to make it fit their narrative.
By mistake, some distortion artists misinterpret outputs, charts or visuals, and communicate what they mean incorrectly. Others willfully distort the meaning of data, and just want the “data science” rubber stamp on something, with or without approval. They might snoop through your work to cherry-pick results they like or, in some cases, replace results with something they would rather show. Either of these are unscientific approaches to the use of data are not appropriate. I find that this persona is rather rare in the wild, but a willful misrepresentation of data is a really difficult and delicate subject to handle.
For the people that unknowingly distort your work, it’s just a matter of patience and being a clear communicator. You may have to put insights in quotes to help them with the wording. I’ve found this to be useful for talking about confusion matrixes or correlation coefficients (concepts that have specific statistical language but can be communicated to a wider audience if handled correctly). For your colleagues that seem desperate to distort your work for personal reasons, be prepared to challenge them on it. Start by speaking with your leader and asking for their guidance. It’s awkward, but it has to be done. In the end, as part of the team conducting the analysis, your co-signature is needed and you need it to be known that you aren’t co-signing if you don’t feel comfortable. Frame this issue of misrepresenting data to leadership as what it is; an ethical concern.
The Reluctant Subject Matter Expert
Part of almost every data science project is learning about the product or process that is meant to be enriched by you being there. This usually means getting teamed up with Subject Matter Experts, or SMEs, to help you understand the current state. Location of relevant data, filters and business logic applied to data are some examples of key items that the data scientist will need to know, and what the SME is meant to provide. In some projects, this person’s knowledge or approval is a straight up requirement for success. That’s what makes it so difficult to work with a SME that only begrudgingly helps out, and only on their time.
Classic reluctant SMEs are simply “too busy” to be spending time with data scientists on projects and making the time by talking with their leadership is off the table. (If an SME truly doesn’t have any time because their leadership doesn’t see value in Data Science, then that’s a whole different story). Their collaboration with data scientists may be limited to pointing to a database, but no further information provided on the tables or contents.
Other reasons for reluctance to help usually stem from some skepticism or fear of any change to their product or process. It’s also possibly that this reluctance comes from a fear of automation.
Talk about how the improvements of a business process can free up time for people to concentrate on more complex problems. Show an example of how automation lets analysts do more analysis, versus repetitive processes to acquire data. If some bureaucratic nightmare is the reason why an SME can’t help you, don’t worry, it’s not the job of the data scientist to overcome that obstacle for everyone. Let people know how crucial the SME can be to success and how the project is at risk without their full support.
Conclusion
While there are some difficult personas to work with, not all is lost. You have options for what you can do to make your workplace more conducive to data science. Look out for my next article on how data scientists can be difficult to work with.
