 Perfect. This is actually, I'm Selena and this is actually very frightening for me because I'm a nodder and I rely on nodders throughout a talk. So I will be keeping an eye on a chat and I'm gonna assume emojis are the nods of virtual conference. So here we go. So decision making and successful data analysis. And why did I put the word successful in quotation is because I wanted to set the premise of the talk in questioning what makes an analysis successful and chewing that topic a little bit more. See, I'm totally seeing the emojis and I totally get that. And I'm also gonna pretend all my jokes are landing so there's that part. So what makes an analysis successful? And one of the things that came into mind was actually a Netflix challenge in 2009. Netflix awarded $1 million to the algorithm that has the most increase in the recommendation engine. So that was actually a 10% increase. So they awarded $1 million to that. However, in production, they actually never used it. They actually used in other algorithms that was only 8.43% and they were talking about and in their blog post, they justified that in saying the engineering cost was too high and it wasn't really worth the production to put it into production for that 1.5%. And other thing I want everyone to keep in mind for the next stage of the talk is that Netflix at that time was also shifting their business model. So they were going from rental to streaming which made a difference in their decision making in terms of what algorithm to actually put in production. So on that note then, what makes that successful? Which one was the successful one? The one actually put in production or the one that has the actual most accuracy? And the answer is it depends and it depends on the audience, it depends on who. And one of the quote I really like is from Roger Pang which is, a data analysis is successful if the audience to which it is presented accepts the results. So the core premise of this is that which it means that as the analyst or as the product creator, we're always doing something for an audience even if it's just for ourselves. And in order for it to be considered successful we need to keep that audience in mind. And that kind of follows through what earlier this morning, Angela and Catherine talked about in terms of empathy and understanding users needs and really understanding where the struggle is and where things are. I love the emojis, so keep them coming out. And so actually there's actually a lot of great resources on how to do it. These are the ones that I take inspiration to. The slides will be shared later so don't need to take note of anything. But so I'm actually not here to talk about the how to do it or the what of it but I'm actually here to talk about the what ifs, the why. And what that really means is that I'm here to talk about ethics and I wanna dive that in a little bit. I'm gonna try to do that in 10 minutes or less because I'm not quite sure what my time is and how I'm doing but let's go into it. I know I chose a meaty topic for a 20 minute talk. And I want to ask what are the implications when we really try to design and understand a user and cater that and filter that lens so that our analysis will be successful because as analysts, we do want to deliver, we do want to deliver value in that way. And one of the key things is that our analytical choices throughout the data analysis process is actually being influenced by that factor. Either we're doing it explicitly and we're aware of it or implicitly that we are not aware of it. And because when we're talking about the audience and it comes to the dynamics and data analysis, what we're really talking about is people and relationships. We're talking about the role, your role, your team's role, who your audience is, is there a trust factor between you and your client or your stakeholders? What kind of relationship you have with each other? Is there competing interests? What is the culture of either your organization or even the industry you're in? And when you couple with that, with your internal processes that we may or may not be fully aware of, such as experience or knowledge or skills, our understanding of what the problem we're trying to solve and there's preference to in terms of both in techniques and skills that we're using and all these internal factors that's going on and they flow throughout this entire data analysis process in a way that actually drives and affects how we make decisions throughout this. So it's actually a nice segue from the previous talk which is we wonder why reproducibility is not trivial because the whole process is not trivial. And so you take this really simplified version of the data analysis process or cycle and we go through these processes where we ask these questions and we make these decisions throughout this process. And even though it seems that it feels like that we are just doing a data analysis, we grab the data, we talk to the people, we shape it, we form it, these are all actually peppered with the people factor and our internal that really impacts off throughout it. And these are influences that drives not only what we ask throughout the data analysis but how we ask what we ask. And one of the things that I encourage analysts to when they're doing analysis to ask themselves too is really think about when do you stop your analysis? When do you think, okay, I'm done. And also why did you stop there? And I think that's quite telling. I personally question myself that a lot too when I do an analysis. And we either do this explicitly or implicitly or either we know that we're doing it or we just don't, but it's behind a scene somewhere hidden. And in the end, what data analysis looks like is actually a roadmap of all our decision-making or internal processes. Thank you, I love my visuals too. Couple with our understanding of our subjective world, how we understand our world and also our interactions and relationships with that and all together forms the actual data product, the analysis product that we're creating. And if at this point you're wondering, okay, so what? Like, what's the difference between me writing of a report and communicating it? And the biggest ethical concern I'm proposing here is that data looks objective and it feels authoritative when it is not, when in fact, especially when we are trying to understand and develop a analysis that really delivers value for a certain audience, the more we do that they're actually less objective, it really is. And we're actually filtering an analysis and data through a very specific lens. And we take the subjective process when we actually create some kind of product and in the end it's being received and perceives as objective. There is an attraction, there's a natural attraction to seeing data in its abstract, being abstract away and boiled down to simplicity because that's tangible and that's what we can measure kind of like our heart rate and an a fifth bit, like does it really, how does it really affect health? Is this something we're really measuring? Why are we measuring what we're measuring? And there's this quote that Philip E. Agro wrote that technology at present is covert philosophy. And I wanna borrow that and say that analysis is covert ideology. It contains a lot of the analysts and all the people involved in terms of their belief system, their thoughts, their biases, it may be to an extent there's personality too. And what that means is that as the person who creates these analysis, these first people who create these data products, even when you're not physically sitting at the table where decisions are being made, a part of you is there in the form of a data analysis. And you are there. And the argument that you're just an engineer, I'm just crunching the numbers, all these, you're never really just an engineer or any of these roles you're playing because we can't separate you from your processes. And all those decisions you made previously are all embedded into your final product. And this quote, objectivity is a subject's delusion that observing can be done without him. Involving objectivity is abrogating responsibility hence its popularity. When we call our data, even down to a data point to a flow-blown analysis when we try to call it objective, we're actually doing a disservice because especially for those who are not familiar with data and not being used to what data really means, they really take that for a grain of salt and take that as granted and be like, yeah, that's what the numbers are telling me. And I'm not here to say that we should just stop designing. I don't think so. I think that there's importance in understanding and delivering value. What I'm trying to advocate more is have that analytical self-awareness that we are doing it. We are actively being involved in it and we shouldn't just step away and call it being like, oh, I'm just punching in the numbers. And there's actually a second part to his statement which is the point is to make it openly philosophical. And I also think that I also want to borrow this and to make the point is to make our decision process openly an actual decision process as opposed to a subconscious way where like, yeah, I'm just shifting this data here and filtering those ones out and being open to that. And as part of a call to action, really, I don't really have none. I think this is a very complex problem. I don't have an answer to it and what I really want to do is actually start some kind of open discussion in terms of how we address this because whether we are aware of it or not, we like it or not and admit it or not, these are things behind the scenes that happening all around in terms of our analysis and being in our data role. And how do we go about in addressing that while delivering value? And in this book that I read before, Mike also argues that as data creator or tech creators, they're the gatekeepers and they're received to an extent they're responsible to do for their products being put out to it. And as much as I enjoy the book, I had a hard time with the term gatekeeper because that sentiment implies power, like that I have power over someone as the analyst and some kind of us versus them mentality, like we're here to stop that out of gate. But in fact, I think all the stakeholders should stay, it's a collaborative effort and it requires everyone at the table to be involved in that process. So I'm not sure if gatekeeper is the best way to address that. So what I'm trying to advocate more is intentionality and inclusivity. By intentionality, I mean really being intentional and involving people to be the intentional that what we're doing is the tradeoff, what we're doing is a decision we're making. Every data point that's filtered out is there's a reason for that and documenting that. Steward's a good one too, yeah. And inclusivity in a sense that get as many people as you can at the table to talk about this as a check and balance with our thoughts and our processes because everyone has a different relationship with people and their role and their world and their understanding. So I really think openness and inclusivity is what really is gonna drive that and it's gonna be challenging because the more difference there are, the more we can, of course my dog barks now too. And I lost track of my thoughts. There you go. And what I wanted to say is and not to be super meta but I am also influenced by so many, so many wonderful people in the data community as well as in my personal life and I would like to give them a thank you and shout out for all of them because they clearly influence how I was, I'm thinking this problem and it's not just something out of my brain. And this is me, feel free to ask questions here in the Slack or Twitter and that is it. I'm gonna clap for myself because I heard someone say big round of applause and I'm awkward that way so. Woo! Good job. Woo! We do have a little bit of time for questions. So I'll open up the floor. If anyone is keen on asking Selena any questions about her talk, like in general. And I kept thinking about Gabrielle's talk as I was there in session four. I really highly recommend for those who missed out on that to revisit that in the cultural meaning of programming language because I think there's a bit of that playing role in how we choose what language and how that impacts throughout it. Again, it's not like I have a clear answer for it but I do think that that's a really worthwhile topic to really bring into awareness to discuss. Give them some good feedback there. Thank you.