In today's world, getting tech talent and bringing them on board is one big challenge for companies

BarRaiser

In 2017, my partner and I launched the first mobile app for children. I took the position of CEO. Because of the lack of a lot of relevant experience, I began to make mistakes almost from the very beginning. I didn’t really know how to “swim,” - yet, I suddenly found myself in the open sea. Nevertheless, even back then, I understood that it is impossible to build a business completely without mistakes. Yes, I did a lot of things, but every time it gave me experience, which I invested in the further development of the project. Let me introduce myself. My name is Mikhail Kotlov. I am a co-founder of , a start-up build around educational applications for children that have been translated into more than 30 languages. Our apps are in the top 10 in the Apple App Store (Kids 5 & Under category) in 40 countries worldwide. We are mainly focused on the European and United States markets, now launching in China. Our product line includes 3 children's applications, an app for parents, a project with online lessons, and an educational animated series "Team Mendeleev". IntellectoKids Today I will share the mistakes that I made while building IntellectoKids, and most importantly, the lessons I learned in the process. Started with the "discovery of America" For many startups, the United States is the most desirable market and main target. I was no exception. Launching IntellectoKids, we thought ambitiously and decided to immediately promote in the United States. In our case, this turned out to be a mistake because advertising in America was more expensive than we expected when planning the budget. In addition, at that time, we did not have enough competencies for successful promotion. Simply put, we weren't ready - and we could do it. Therefore, we quickly changed our strategy and turned to less competitive markets. We have localized the product into Swedish, Spanish, and French markets. It worked: the company began to grow. Of course, then we returned to the USA, but now we had more experience and money, so everything turned out well. Now our applications are very popular in the States. So if you have already tried to storm the US market and failed, do not worry: look at the countries of Latin America and Europe. Perhaps, you will have a better chance there. Locked all processes on me When you only have 8-10 employees, you are thoroughly immersed in everything that happens in your startup. Almost no phone call in the team is complete without your participation. This is not uncommon, and in the beginning, this is the right strategy. However, at some point, the company was already growing in terms of metrics, the team was getting larger, the processes were growing, but the approach stayed the same. This is what was happening at IntellectoKids. I was aware of the smallest details: any problem, any solution passed through me. As a result, I had practically no time left for strategic planning. And the CEO's task is to constantly check the course, anticipating everything several steps ahead. It's good that we have two founders at IntellectoKids, and my partner partially covered this problem. Immersion in routine processes interferes with researching trends and communicating with people from the field. However, this is necessary to keep the pulse on market trends and news. Eventually, new ideas come less often because you are too busy with operations and troubleshooting every small issue. Unintentionally, my behavior looked like distrust for my team. It partly prevented the company's team leads from developing professionally. Until I dared to trust my team, they entirely relied on my opinion and could not become independent "players". As a result: It started hurting business goals; It made the team bored and burned them out Fortunately, it didn’t come to a critical point, and I got it together. We now call the team leaders once a week to discuss global product and marketing issues and adjust the action plan if required. Plus, I do regular one-on-one calls to make sure things are going as planned. This tactic saves resources and avoids distraction through trivial issues. As it turns out, my team is capable of brainstorming ad creatives or controlling the release of a new version of the application without me. Today, the creative team and the development team are now working independently. I would like the same thing to happen over time with all teams, but it takes time to cultivate talent and find candidates with leadership skills in the job market. For example, currently, we are now looking for a competent CMO. As of right now, I take care of the CMO duties as well. Worked with digital agencies There was a time when I thought it was easier to outsource marketing and creativity. I was wrong. Although for many companies the agency is a great option, it turned out that in my case, it was not suitable. The reason is the particular aspects of our niche market. It is difficult to advertise an app for preschoolers. First, you have two target audiences: children and parents. The child needs an exciting product, and the parent needs a useful one. Therefore, it is important to show the educational value of the app, not just the demo of the game. The second difficulty is that parents have a very wide range of interests. When we advertise a children's app, we compete with all products and services of interest to women between the ages of 25 to 55 - this can be anything from match 3 games to beauty salons. We must catch moms' attention in a limited period of time while they are on Facebook. The field of children's applications is quite young. Most agencies do not have a sufficient understanding of how to work in this field. It is very difficult to gain experience in conducting such advertising campaigns if you work in an agency and are not thoroughly immersed in the project, like the guys at IntellectoKids. Therefore, at some point, I had to abandon the idea of outsourcing. I've built an in-house team of creators and marketers who do a better job than agencies. Didn't share ideas "Sitting inside a box" was a big mistake on my part. I did not talk about projects and plans. I was a little afraid that my ideas would be stolen and was superstitious: I did not want to announce any new features until they became a reality. In fact, you are unlikely to come up with something completely new. In other words, there is nothing new under the Sun. So you shouldn't be afraid of collaborations or discussions. By staying silent about your ideas, you rob yourself of the opportunity to get useful feedback. I think I missed a lot of useful comments from smart and experienced people who, at that time, understood business processes better than I did. Moreover, if you don’t share ideas and plans, you will never learn how to make a great product or promote it. In the long run, exchanging ideas and feedback is a lot more valuable than staying silent. Finally, there was a need to raise capital, and I realized that it is wrong to stay silent. One of the founders' responsibilities is to get the attention of potential investors. It means constantly telling people about the company. You never know which contact will invest or direct you to the right investor, so you need to keep your business network informed and excited about your company. When asked for advice, I always say: "Be generous." Share your ideas and news. The established connections will definitely come in handy one day. Didn't develop a recruiting team For three years, I had just one recruiter, and we built up the marketing, development, and product teams together. At the time, I didn’t think about the need to build a recruiting team to optimize the hiring process. The Russian market is not the target market for IntellectoKids; we work little with this category of users. People are not familiar with our product, and practically, not familiar with the company. The fact that we are not heard of significantly narrows our opportunities for recruiting specialists and reduces the number of applications for vacancies. To increase the company's recognition in the Russian job market, we began to engage in PR. Currently, we are working to create our HR brand. But it should not stop there. It's really great in our company: we create high-quality content, work with modern development technologies, offer various pleasant bonuses - from voluntary medical insurance to pay for specialized training. But someone has to “sell” all this to candidates, hunt people from well-known companies, and use non-standard search methods. In this case, you cannot do without a strong recruitment team, so I recently allocated resources to create a recruiting department. Interviewed people incorrectly I didn't always know how to conduct interviews. I thought I was asking the right questions. But in fact, I missed the important details, which subsequently influenced the person's work and his relationships with other team members. These details were mostly values ​​- and if a candidate’s value matched the company values. It is much easier to assess professional skills in a regular interview than soft skills. And it can be difficult to understand whether a newcomer can organically integrate into the team. The simplest example: once, I hired a girl who, as it turned out, was used to swearing at her colleagues. This is not accepted in the culture of IntellectoKids: it’s crazy for us to even raise our voices. Unfortunately, the conversations did not affect, and we had to part ways with that employee. And I had to continue to search for a suitable person for her role. Now I stick to this rule: during interviews, I pay attention to soft skills. If a candidate doesn’t check all the boxes on soft skills, I won't hire this candidate, even if he/she is a genius. After a couple of unfortunate incidents, I decided to dive deeper into the issue and look for a system that would help me hire competent people. One of my acquaintances recommended “The Who method”, as described in the book by Jeff Smart and Randy Street. (See, not being silent and talking to people helps!) The selection process is quite long. On average, it takes two to three weeks from the first interview to an offer. But the method works: during the time that we have been using it, there have been almost no cases in IntellectoKids that a person did not pass the probationary period. So, the system is useful for everyone involved. How does it work? First, the interview is conducted by the recruiter to check whether the person is generally suitable. For example, we check if the candidate’s goals align with the company's goals. Then, we give a test task. With the test task, we obviously check the hard skills. However, we can see how this candidate thinks and how he/she can immerse themselves in a topic. The next step: the team-lead interviews the candidate to see if the candidate has professional skills. The next step is an interview with me. If the candidate is not on my team, I act as a barraiser. Barraiser is an expert invited for an interview who is not directly interested in the candidate. The interview is all about identifying the strengths and weaknesses of soft skills. This is done, among other things, so that both parties understand how comfortable it would be to work together. If the conversation went well, we would collect references from past jobs. Candidates are often intimidated by this point. However, there is no reason to be scared here. We might not extend an offer because of negative recommendations, only when something truly terrible that this candidate has done in their previous job. Finally, after this stage, we extend an offer. It turns out that getting to us is quite difficult. But we have a low turnover rate. Usually, each hire is a win-win match. The main thing is not to be lazy and not to skip any of the selection steps. There were a few times when we didn’t follow each step in our hiring method. As a result, the newcomer either left or got fired. Not enough investment in server development Until recently, our server development was based on part-time employees. We did not have a full-fledged team in this direction. Meanwhile, server-side development is very important for monetization, analytics, and marketing, because all analytical and marketing tools require server-side solutions. Otherwise, it is impossible to draw the right conclusions, analyze the data and understand what is more profitable for you to sell and how to promote the product. The lack of a competent full-time server development team began to hinder the development of our marketing and analytics. And in the end, when it was necessary to grow, we were engaged in the development of server architecture. Actually, we are still fixing this mistake. Stopped running CustDev I started my business researching parenting pain. But at a certain stage of company growth, I stopped further researching on the subject. It was definitely my fault. CustDev must be persistent. First, the needs of your target audience are always changing, and you need to improve your content in accordance with those new needs. Secondly, your product changes, and you need to understand if you are on the right track. We recently started to focus on CustDev again, but due to lack of time, we are not as active as we would like to be. Got stuck on Facebook We researched different promotion channels; our main focus was Facebook. Our ads worked so well; we were tempted to spend our entire marketing budget on Facebook. But through practice, we learned that it wasn’t the best approach. Even when you find the most profitable channel for your product, it is important to use other channels as well. Currently, we are looking for new team members who will be engaged in promotion in Google Ads, influence marketing, CPA networks, and different partnerships. I wanted to get results without investing in the team When we launched online lessons, we needed to look for teachers. At first, I took care of it myself. I looked for candidates myself and attracted the head of the online lessons department. However, real results started to happen when we hired a dedicated recruiter with the appropriate expertise. There are many similar examples in my practice. As a result, I strongly believe in hiring the right people for the task. Instead of dividing new responsibilities between existing employees, who may or may not have enough competencies to succeed with this particular task. If you are trying to reduce financial risks without investing in people, you are still taking risks, the risk of undesirable outcomes. When you get these outcomes, you will think that the reason is in the product itself and in the lack of demand. The chances are that your company simply does not have the expertise that would allow the product to grow. Or maybe your current team doesn't have enough time to devote to the project. In both cases, the solution is to hire people who will only focus on the new task. Staked on game content We develop educational games for children. Previously, the process of creating content was structured as follows: first, the game designer came up with a scenario for the game, and then we "built" learning mechanics into it. During the pandemic, the demand for systematic education of children through applications has grown. The parents wanted the child to not just play and, perhaps, learn a new skill as a bonus. Our target audience needed something more: a product that carries an understandable educational value, and only then, just as a bonus, entertains. Therefore, our apps now help children prepare for school through gamification. First, the teaching methodologist writes an educational program, and only after that, the game designer packs it up and makes an exciting game. This point can be attributed to errors only conditionally since we rebuilt the processes in time. Still, we could have done this earlier; it would only help us gain more users. How I learned and still learning from my mistakes To be wrong in the “right” way, I carefully analyze past experiences and understand: what exactly went wrong; what risks in the future does a particular mistake carry; what opportunities it opens up for me; what can I fix - and how; how can I minimize losses at the same time; what needs to be done so that the mistake does not repeat itself. Also, I advise everyone to read the book by Carol Dweck, “Flexible Mind.” The author of the book, a professor at Stanford University, explores the nature of the fear of mistakes and gives advice on how not to be afraid of new experiences. The book also talks about how we can only learn through our mistakes.

Mistakes I Made While Building an EdTech Startup With 400K Active Users

Much talk has been thrown around regarding the importance of diversity; however, as Dave McClure of 500 Startups noted, In general, that means statistically more “white male nerds” get hired and receive funding Dave said, “There’s a soft bias towards doing things that are familiar.” “ — I think we are making progress,” “Now it’s brown and yellow nerds that get funding.” As a small , your first hire can make or break your team, and as teams grow, there is a tendency to bring on people who remind you of yourself. What are actual ways to make sure you are hiring a diverse team in backgrounds as well as skillsets? startup natural I recently attended the x Women in Tech event with Alexandra Wilkis Wilson (Co-Founder of Glamsquad, previously Co-Founder of Gilt), Sutian Dong (Partner at Female Founders Fund), Kate Shillo (Director at Galvanize Ventures), and Ana Mahoney (Uber General Manager). Here are my actionable takeaways from the event: Uber Vettery Your core attracts your applicants. Take a hard look at your core team The core of your team will inevitably attract similar applicants. People are drawn to people who remind them of themselves. Sutian retells the story of two male co-founders, sons of immigrants who were adamant about ensuring diversity in hiring. Only when they took a company picture at the end of the year, did they realize there were no women in the picture — they had resort to pulling random women to stand in the picture. The lesson? When people look for jobs and think about fit, they’re going to look at the current makeup of the team, and it’s easier to ensure there is diversity early on when there are five people rather than when there are 50. It is difficult to “hire for diversity” when your core itself is not diverse. Adding in a simple sentence at the end of job descriptions that underscores your commitment to diversity, e.g. “we encourage people of all backgrounds, and all experiences to apply,” will change the makeup of applicants. Actionable tip: Consider bringing on a “bar-raiser” for your hiring process The bar raiser is an objective party that does not work for the company and is tasked to ask the questions that the hiring team doesn’t think to ask in order to prevent assumptions or biases that otherwise might be overlooked. The bar raiser should effectively moderate interview debriefs, making sure all concerns are aired and At Uber, Anna stresses, “We hire for teams, and when you have a diverse group of people as part of that process, then you naturally attract diverse applicants, and that mindset is reflected in those we hire.” be able to candidly say, “Here are the criteria we have for hiring, here is what we have, and here is what we’re missing.” Use an assessment test like Myers Briggs to measure diversity of strengths Early on at Gilt, Alexandra brought on an executive coach to build out a highly functional team. At the time a team of seven, they used the Myers Briggs test to assess their profiles. The result? When all team members were mapped out on 16 quadrants, they discovered they were all very similar in how they thought and how they processed information. It’s human tendency to hire people that remind you of yourself, and “It’s important to think about diversity not only in how people look, but also in how they think.” entrepreneurs tend to be “idea people, glass half-full optimists. Make sure to balance this out with others who are are analytical, realistic, and metrics-driven. Test for skills rather than for previous experiences first There’s a natural inclination to ask about someone’s past work experiences. However, at Uber, they test for skills first by administering real life case studies that reflect what you would actually be working on in the role. If you’ve enjoyed doing those assignments — whether that’s structuring a marketing campaign, writing a blog post, organizing operational flow — then both sides actually get a sense of if you would thrive in that role. If you’re on the hiring side, ask these questions: Ask questions about a puzzle you are trying to figure out, and see how they analyze the pieces. “Are they structured in their thinking? Are they articulate? Are they creative? Can they pick up on a hint and build upon it?” Finally, conduct a thorough background check Think you’ve found your dream candidate? Make sure you don’t overlook reference checks. To make sure you are unbiased in hiring, Sutian recommends having your company’s board members meet any executive team hires. Reference check who they’ve worked with before, even check with people on LinkedIn that you are mutually connected with. One challenging question that Alexandra always asks references is References tend to err on reviewing candidates positively, but this question can shed light on potential red flags such as, “x tends to be impatient and had some difficulty working in our small team.” Kate encourages people to “check [the candidate’s] credit, especially if they’re managing your money.” “What’s the worst thing you can say about this person?” At SheWorx , we believe you need to be authentic about your reasoning behind why you believe diversity is important to you. Make sure that if it is a priority, you are clear about that during your interactions with the candidate. If you fundamentally don’t know what your reasoning is behind supporting diversity, then take the time learn the facts, read the stories, and analyze different viewpoints, so you can formulate an opinion that aligns with your values and beliefs. The worst thing you can do is to say you are committed to diversity and not be able to articulate why it’s important to you. When you share your reasoning, cite a story that makes you believe that improving diversity is important. People can smell inauthenticity and that’s the last thing you want to project to a potential hire. SheWorx is a collective of ambitious female entrepreneurs and changemakers redefining the next wave of leadership. RSVP for our next SheWorx Breakfast Wednesday 2/3 with Owen Davis , Founder of Overlap, NYC Seed, and Alpha Ventures. He’ll be providing actionable strategies for how to raise capital efficiently.

How to Build a [Strong and Diverse] Winning Team

Breaking Down Organizational Barriers to Accelerate Software Development

সফ্টওয়্যার বিকাশকে ত্বরান্বিত করতে সাংগঠনিক বাধাগুলি ভেঙে ফেলা

Abbau organisatorischer Barrieren zur Beschleunigung der Softwareentwicklung

Derribar barreras organizacionales para acelerar el desarrollo de software

Supprimer les barrières organisationnelles pour accélérer le développement de logiciels

सॉफ्टवेयर विकास में तेजी लाने के लिए संगठनात्मक बाधाओं को तोड़ना

組織的障壁を打破しソフトウェア開発を加速する

조직의 장벽을 허물어 소프트웨어 개발 가속화

Quebrando barreiras organizacionais para acelerar o desenvolvimento de software

Разрушение организационных барьеров для ускорения разработки программного обеспечения

Yazılım Geliştirmeyi Hızlandırmak için Organizasyonel Engelleri Yıkmak

Phá vỡ các rào cản tổ chức để tăng tốc phát triển phần mềm

打破组织障碍，加速软件开发

Authors: (1) Jonathan H. Rystrøm. Table of Links Abstract and Introduction Previous Literature Methods and Data Results Discussions Conclusions and References A. Validation of Assumptions B. Other Models C. Pre-processing steps B Other Models B.1 Model without activity-level However, when we ran the model (also using lmerTest (Kuznetsova et al., 2017)) and plotted the residuals, we got the following: Just from a brief visual inspection, it is clear to see that the residuals are not randomly distributed: There are two distinct ”bands” that both seem to trend upward. This breaks the assumption that the residuals are randomly distributed (Poole & O’Farrell, 1971). While many assumption violations are reduced with enough data (Baayen et al., 2008; Schielzeth et al., 2020), non-linearity is not one of them (Poole & O’Farrell, 1971). Fortunately, the non-random residuals were (partially) fixed by introducing activity-level for the reasons described in section 3.3. B.2 Problematic Categories Removed Here we fit the main model (eq. 5, with problematic categories removed. We define a problematic category as a category with a fitted random effect of less than -0.5. We obtain this threshold by visually inspecting Fig. 7. The results of this fit can be seen below in 2: B.3 Gamma mixed-effects model In the following, we refit the main model (eq. 5) using a Gamma regression. This is the most widely recommended solution in the literature on fitting right-tailed, heteroscedastic outcomes (Feng et al., 2014; Villadsen & Wulff, 2021). However, since we discovered this after running our initial models, we could only justify doing this as a post-hoc test. I use the lme4-package to fit the model (Bates et al., 2014). To avoid convergence errors and adapt the model to the formula, we make the following alterations: Add a gamma log-link function (Fox, 2015) change the year (β1) estimate to decades. This has the effect of rescaling the effect size. We still log-transform the activity-level to rescale it. As this is not part of the hypothesis, this does not affect our interpretation. This leads us to the interpretation. The effect size per decade is 0.31, which is highly significant (SE=0.006, T=24, p ≪ 0.001). This translates into 34% increase per decade or 3% increase per year. This is the same direction as our transformed linear model, with somewhat larger results. However, as these come from an almost unidentifiable fit with extremely non-random residuals, no inferences can be drawn from this. Some of this indicates that the conditional distribution of Barrier-to-Exit is not a Gamma distribution. Pursuing the GLMM path would require further assessments of the best-fitting distribution. This could be by e.g. applying the Box-Cox method as described by Villadsen and Wulff (2021). This paper is under CC 4.0 license. available on arxiv

Assessing Non-linearities and Distribution Assumptions in Barrier-to-Exit Analysis

Authors: (1) Jonathan H. Rystrøm. Table of Links Abstract and Introduction Previous Literature Methods and Data Results Discussions Conclusions and References A. Validation of Assumptions B. Other Models C. Pre-processing steps A Validation of Assumptions There are several assumptions to validate for crossed-linear mixed effects models (Baayen et al., 2008). Specifically, we need to assess: • : Is the phenomenon the model captures actually linear? This can be tested by looking for trends in the residuals. If no trends are found, linearity holds (Poole & O’Farrell, 1971). Linearity Is the variance across residuals equal? This can be tested by plotting residuals against response and looking for cone-like shapes (Fox, 2015). • Homogeneity of Variance: Are the residuals normally distributed? We can assess this by a QQ-plot over the residuals. • Normality of residuals: Per the definition of crossed-effects mixed effects models (Baayen et al., 2008), we expect the residuals to be normal. This can be assessed similarly to the random effects. • Normality of random effects: Note that testing the assumptions relies on visual tests rather than statistical tests. This is not for a lack of statistical tests (see e.g. Fox, 2015). Rather it is because statistical tests tend to become oversensitive for large datasets (Ghasemi & Zahediasl, 2012). Given our dataset has more than 50,000 observations, it is safer to use visual tests. First, let us test for linearity and heteroscedasticity. The residual plot can be seen in fig. 5: At first glance, there seems to be no strong heteroscedasticity or apparent linearity. However, at a closer look, there are some weird outliers in the bottom left corner. Furthermore, there is a weird kind of straight line going from the top left corner and down towards the middle. This indicates that there may be something wrong with the model. Let us now assess the normality of the residuals: We notice that the residuals are generally lower than the normality line. This indicates a left skew in the residuals, i.e. there are smaller outliers. While the fixed effects are robust against this type of skew, the estimates for group-level variances are more affected (Schielzeth et al., 2020). Finally, let us consider the normality of the random effects. A qq-plot of the random effects for the two levels can be seen below: Here we see a dramatic skew toward the lower level of the fitted residuals. This implies that there are weird dynamics in the lower values of Barrier-to-Exit. Part of this may be because of the relatively few observations per user (see fig 8). This highlights another issue with measuring Barrier-to-Exit across time: because it requires many observations over a long period of time to calculate sufficient precision of revealed preferences, it’s hard to measure changes on the individual level. While the effects of non-normality on the fixed effects should be minor (Schielzeth et al., 2020), we nevertheless conduct a robustness check. In the robustness check, we remove the problematic categories and refit the main model (eq. 5). The results of this can be seen in Appendix B.2 This paper is under CC 4.0 license. available on arxiv

Ensuring Model Validity: A Closer Look at Assumptions in Amazon's Barrier-to-Exit Analysis

Authors: (1) Jonathan H. Rystrøm. Table of Links Abstract and Introduction Previous Literature Methods and Data Results Discussions Conclusions and References A. Validation of Assumptions B. Other Models C. Pre-processing steps 6 Conclusion Understanding how recommender systems shape our behaviour is essential to avoid manipulation. In this paper, we investigated the Amazon recommender system concerning whether it has made it harder to change preferences. By analysing the Barrier-to-Exit (Rakova & Chowdhury, 2019) of more than 50,000 users, we found a highly significant growth in Barrier-to-Exit over time, which indicates that it has indeed become harder to change preferences for the analysed users. However, sampling bias induced by the calculation of Barrier-to-Exit makes it difficult to draw conclusions about the general population of Amazon customers. This highlights the dilemma of portability in measuring socio-technical systems (Selbst et al., 2019): accurately evaluating a concept like ”changing preferences” requires adapting to the context of the system, which makes it more difficult to generalise (and compare) to other systems. Comparing recommender systems is necessary for ensuring that these respect human autonomy (Varshney, 2020) and live up to new regulations such as the EU AI Act (Kop, 2021). Further work, should aim to create auditing procedures and metrics that allow third parties to measure potential preference manipulation in a way that fits within the context of the industry and allows for comparisons between different systems. This will help assess the pressures of Surveillance Capitalism (Zuboff, 2019) on human autonomy. References Baayen, R. H., Davidson, D. J., & Bates, D. M. (2008). Mixed-effects modeling with crossed random effects for subjects and items. Journal of memory and language, 59 (4), 390–412. (Publisher: Elsevier) Barto´n, K. (2022, September). MuMIn: Multi-Model Inference. Retrieved 2023-01-08, from https://CRAN.R-project .org/package=MuMIn Bates, D., M¨achler, M., Bolker, B., & Walker, S. (2014, June). Fitting Linear Mixed-Effects Models using lme4. arXiv. Retrieved 2023-01-12, from http://arxiv.org/abs/1406.5823 (arXiv:1406.5823 [stat]) Bennett, J., & Lanning, S. (2007). The netflix prize. In Proceedings of KDD cup and workshop (Vol. 2007, p. 35). New York. urrell, J. (2016, June). How the machine ‘thinks’: Understanding opacity in machine learning algorithms. Big Data & Society, 3 (1), 2053951715622512. Retrieved 2022-10-11, from https://ezproxy-prd.bodleian.ox.ac.uk:2246/doi/full/10.1177/2053951715622512 (Publisher: SAGE Publications Ltd) doi: 10.1177/ 2053951715622512 Calvo, R. A., Peters, D., Vold, K., & Ryan, R. M. (2020). Supporting Human Autonomy in AI Systems: A Framework for Ethical Enquiry. In C. Burr & L. Floridi (Eds.), Ethics of Digital Well-Being: A Multidisciplinary Approach (pp. 31–54). Cham: Springer International Publishing. Retrieved 2022-08-16, from https://doi.org/10.1007/ 978-3-030-50585-1 2 doi: 10.1007/978-3-030-50585-1 2 Carroll, M. D., Dragan, A., Russell, S., & Hadfield-Menell, D. (2022). Estimating and penalizing induced preference shifts in recommender systems. In International Conference on Machine Learning (pp. 2686–2708). PMLR. Feng, C., Wang, H., Lu, N., Chen, T., He, H., Lu, Y., & Tu, X. M. (2014, April). Log-transformation and its implications for data analysis. Shanghai Archives of Psychiatry, 26 (2), 105–109. Retrieved 2023-01-10, from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4120293/ doi: 10.3969/j.issn.1002-0829.2014.02.009 Floridi, L., Holweg, M., Taddeo, M., Amaya Silva, J., M¨okander, J., & Wen, Y. (2022, March). capAI - A Procedure for Conducting Conformity Assessment of AI Systems in Line with the EU Artificial Intelligence Act (SSRN Scholarly Paper No. ID 4064091). Rochester, NY: Social Science Research Network. Retrieved 2022-04-01, from https://papers.ssrn.com/abstract=4064091 doi: 10.2139/ssrn.4064091 Fox, J. (2003). Effect Displays in R for Generalised Linear Models. Journal of Statistical Software, 8 (15). Retrieved 2023-01-10, from http://www.jstatsoft.org/v08/i15/ doi: 10.18637/jss.v008.i15 Fox, J. (2015). Applied regression analysis and generalized linear models. Sage Publications. Franklin, M., Ashton, H., Gorman, R., & Armstrong, S. (2022, May). Missing Mechanisms of Manipulation in the EU AI Act. The International FLAIRS Conference Proceedings, 35 . Retrieved 2022-12-31, from https:// journals.flvc.org/FLAIRS/article/view/130723 doi: 10.32473/flairs.v35i.130723 Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., Daum´e III, H., & Crawford, K. (2021, December). Datasheets for Datasets. arXiv. Retrieved 2022-10-20, from http://arxiv.org/abs/1803.09010 (arXiv:1803.09010 [cs]) Ghasemi, A., & Zahediasl, S. (2012, December). Normality Tests for Statistical Analysis: A Guide for NonStatisticians. International Journal of Endocrinology and Metabolism, 10 (2), 486–489. Retrieved 2023-01-11, from https://brief.land/ijem/articles/71904.html doi: 10.5812/ijem.3505 Gorelick, M., & Ozsvald, I. (2020). High performance Python: practical performant programming for humans (Second edition ed.). Beijing [China] ; Boston [MA]: O’Reilly. Harper, F. M., & Konstan, J. A. (2016, January). The MovieLens Datasets: History and Context. ACM Transactions on Interactive Intelligent Systems, 5 (4), 1–19. Retrieved 2022-12-29, from https://dl.acm.org/doi/10.1145/ 2827872 doi: 10.1145/2827872 Harris, C. R., Millman, K. J., van der Walt, S. J., Gommers, R., Virtanen, P., Cournapeau, D., . . . Oliphant, T. E. (2020, September). Array programming with NumPy. Nature, 585 (7825), 357–362. Retrieved 2021-12-27, from https://www.nature.com/articles/s41586-020-2649-2 doi: 10.1038/s41586-020-2649-2 Jannach, D., & Adomavicius, G. (2016, September). Recommendations with a Purpose. In Proceedings of the 10th ACM Conference on Recommender Systems (pp. 7–10). Boston Massachusetts USA: ACM. Retrieved 2022-09-01, from https://dl.acm.org/doi/10.1145/2959100.2959186 doi: 10.1145/2959100.2959186 Jiang, R., Chiappa, S., Lattimore, T., Gy¨orgy, A., & Kohli, P. (2019, January). Degenerate Feedback Loops in Recommender Systems. In Proceedings of the 2019 AAAI/ACM Conference on AI, Ethics, and Society (pp. 383–390). Honolulu HI USA: ACM. Retrieved 2022-09-07, from https://dl.acm.org/doi/10.1145/3306618 .3314288 doi: 10.1145/3306618.3314288 Kipf, T. N., & Welling, M. (2017, February). Semi-Supervised Classification with Graph Convolutional Networks. arXiv. Retrieved 2022-12-14, from http://arxiv.org/abs/1609.02907 (arXiv:1609.02907 [cs, stat]) Knijnenburg, B. P., Reijmer, N. J., & Willemsen, M. C. (2011, October). Each to his own: how different users call for different interaction methods in recommender systems. In Proceedings of the fifth ACM conference on Recommender systems (pp. 141–148). New York, NY, USA: Association for Computing Machinery. Retrieved 2023-01-11, from https://doi.org/10.1145/2043932.2043960 doi: 10.1145/2043932.2043960 Kop, M. (2021, November). EU Artificial Intelligence Act: The European Approach to AI. , 11. Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest package: tests in linear mixed effects models. Journal of statistical software, 82 (13). (Publisher: The Foundation for Open Access Statistics) Lam, S. K., Pitrou, A., & Seibert, S. (2015). Numba: A llvm-based python jit compiler. In Proceedings of the Second Workshop on the LLVM Compiler Infrastructure in HPC (pp. 1–6). Ledwich, M., & Zaitsev, A. (2019). Algorithmic extremism: Examining YouTube’s rabbit hole of radicalization. arXiv preprint arXiv:1912.11211 . Leino, J., & R¨aih¨a, K.-J. (2007, October). Case amazon: ratings and reviews as part of recommendations. In Proceedings of the 2007 ACM conference on Recommender systems (pp. 137–140). New York, NY, USA: Association for Computing Machinery. Retrieved 2023-01-08, from https://doi.org/10.1145/1297231.1297255 doi: 10.1145/1297231.1297255 Linden, G., Smith, B., & York, J. (2003). Amazon. com recommendations: Item-to-item collaborative filtering. IEEE Internet computing, 7 (1), 76–80. (Publisher: Ieee) Maddala, G. S. (1971, March). The Use of Variance Components Models in Pooling Cross Section and Time Series Data. Econometrica, 39 (2), 341. Retrieved 2023-01-11, from https://www.jstor.org/stable/1913349 ?origin=crossref doi: 10.2307/1913349 McAuley, J., Targett, C., Shi, Q., & Hengel, A. v. d. (2015, June). Image-based Recommendations on Styles and Substitutes. arXiv. Retrieved 2022-12-29, from http://arxiv.org/abs/1506.04757 (arXiv:1506.04757 [cs]) McKinney, W. (2011). pandas: a foundational Python library for data analysis and statistics. Python for high performance and scientific computing, 14 (9), 1–9. (Publisher: Seattle) Millecamp, M., Htun, N. N., Jin, Y., & Verbert, K. (2018). Controlling spotify recommendations: effects of personal characteristics on music recommender user interfaces. In Proceedings of the 26th Conference on user modeling, adaptation and personalization (pp. 101–109). Nakagawa, S., Johnson, P. C., & Schielzeth, H. (2017). The coefficient of determination R 2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded. Journal of the Royal Society Interface, 14 (134), 20170213. (Publisher: The Royal Society) Nguyen, T. T., Hui, P.-M., Harper, F. M., Terveen, L., & Konstan, J. A. (2014). Exploring the filter bubble: the effect of using recommender systems on content diversity. In Proceedings of the 23rd international conference on World wide web - WWW ’14 (pp. 677–686). Seoul, Korea: ACM Press. Retrieved 2022-12-31, from http:// dl.acm.org/citation.cfm?doid=2566486.2568012 doi: 10.1145/2566486.2568012 Ni, J., Li, J., & McAuley, J. (2019). Justifying recommendations using distantly-labeled reviews and fine-grained aspects. In Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (EMNLP-IJCNLP) (pp. 188–197). Nonnecke, B., & Preece, J. (2000). Lurker demographics: Counting the silent. In Proceedings of the SIGCHI conference on Human factors in computing systems (pp. 73–80). O’Hara, R., & Kotze, J. (2010, January). Do not log-transform count data. Nature Precedings. Retrieved 2023-01-11, from https://www.nature.com/articles/npre.2010.4136.1 doi: 10.1038/npre.2010.4136.1 Pandita, R. (2017). Internet: A change agent an overview of internet penetration & growth across the world. International Journal of Information Dissemination and Technology, 7 (2), 83–91. Retrieved 2023-01-11, from https://www.proquest.com/docview/1920419032/abstract/14B253C0DD6A4277PQ/1 (Num Pages: 83-91 Place: Ambala, India Publisher: Maharishi Markandehwar University) Papakyriakopoulos, O., Serrano, J. C. M., & Hegelich, S. (2020). Political communication on social media: A tale of hyperactive users and bias in recommender systems. Online Social Networks and Media, 15 , 100058. (Publisher: Elsevier) Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., . . . Dubourg, V. (2011). Scikit-learn: Machine learning in Python. the Journal of machine Learning research, 12 , 2825–2830. (Publisher: JMLR. org) Poole, M. A., & O’Farrell, P. N. (1971). The assumptions of the linear regression model. Transactions of the Institute of British Geographers, 145–158. (Publisher: JSTOR) Popper, K. R. (1970). Normal science and its dangers. Cambridge University Press Cambridge. Rakova, B., & Chowdhury, R. (2019, September). Human self-determination within algorithmic sociotechnical systems. arXiv. Retrieved 2022-09-07, from http://arxiv.org/abs/1909.06713 (arXiv:1909.06713 [cs]) Raschka, S., Patterson, J., & Nolet, C. (2020). Machine learning in python: Main developments and technology trends in data science, machine learning, and artificial intelligence. Information, 11 (4), 193. (Publisher: Multidisciplinary Digital Publishing Institute) Roski, J., Maier, E. J., Vigilante, K., Kane, E. A., & Matheny, M. E. (2021). Enhancing trust in AI through industry self-governance. Journal of the American Medical Informatics Association, 28 (7), 1582–1590. (Publisher: Oxford University Press) Sandvig, C., Hamilton, K., Karahalios, K., & Langbort, C. (2014). Auditing algorithms: Research methods for detecting discrimination on internet platforms. Data and discrimination: converting critical concerns into productive inquiry, 22 , 4349–4357. Satterthwaite, F. E. (1946, December). An Approximate Distribution of Estimates of Variance Components. Biometrics Bulletin, 2 (6), 110. Retrieved 2023-01-11, from https://www.jstor.org/stable/10.2307/3002019 ?origin=crossref doi: 10.2307/3002019 Schielzeth, H., Dingemanse, N. J., Nakagawa, S., Westneat, D. F., Allegue, H., Teplitsky, C., . . . Araya-Ajoy, Y. G. (2020, September). Robustness of linear mixed-effects models to violations of distributional assumptions. Methods in Ecology and Evolution, 11 (9), 1141–1152. Retrieved 2022-12-27, from https://onlinelibrary.wiley.com/ doi/10.1111/2041-210X.13434 doi: 10.1111/2041-210X.13434 Selbst, A. D., Boyd, D., Friedler, S. A., Venkatasubramanian, S., & Vertesi, J. (2019, January). Fairness and Abstraction in Sociotechnical Systems. In Proceedings of the Conference on Fairness, Accountability, and Transparency (pp. 59–68). Atlanta GA USA: ACM. Retrieved 2022-09-07, from https://dl.acm.org/doi/10.1145/3287560.3287598 doi: 10.1145/3287560.3287598 Smith, B., & Linden, G. (2017, May). Two Decades of Recommender Systems at Amazon.com. IEEE Internet Computing, 21 (3), 12–18. Retrieved 2022-12-20, from http://ieeexplore.ieee.org/document/7927889/ doi: 10.1109/MIC.2017.72 Timmers, H., & Wagenaar, W. A. (1977, November). Inverse statistics and misperception of exponential growth. Perception & Psychophysics, 21 (6), 558–562. Retrieved 2023-01-10, from https://doi.org/10.3758/BF03198737 doi: 10.3758/BF03198737 Van Rossum, G. (2007). Python Programming language. In USENIX annual technical conference (Vol. 41, pp. 1–36). (Issue: 1) Villadsen, A. R., & Wulff, J. N. (2021, July). Statistical Myths About Log-Transformed Dependent Variables and How to Better Estimate Exponential Models. British Journal of Management, 32 (3), 779–796. Retrieved 2023-01-12, from https://onlinelibrary.wiley.com/doi/10.1111/1467-8551.12431 doi: 10.1111/1467-8551.12431 Wells, J. R., Danskin, G., & Ellsworth, G. (2018). Amazon. com, 2018. Harvard Business School Case Study(716-402). Wood, S. N., Pya, N., & S¨afken, B. (2016, October). Smoothing Parameter and Model Selection for General Smooth Models. Journal of the American Statistical Association, 111 (516), 1548–1563. Retrieved 2023- 01-12, from https://doi.org/10.1080/01621459.2016.1180986 (Publisher: Taylor & Francis eprint: https://doi.org/10.1080/01621459.2016.1180986) doi: 10.1080/01621459.2016.1180986 Zhu, F., & Liu, Q. (2018). Competing with complementors: An empirical look at Amazon.com. Strategic Management Journal, 39 (10), 2618–2642. Retrieved 2023-01-11, from https://onlinelibrary.wiley.com/doi/ abs/10.1002/smj.2932 ( eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/smj.2932) doi: 10.1002/ smj.2932 Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power: Barack Obama’s Books of 2019. Profile Books. This paper is under CC 4.0 license. available on arxiv

Ensuring Ethical AI: Lessons from Amazon's Barrier-to-Exit Analysis

Authors:
(1) Jonathan H. Rystrøm. Table of Links Abstract and Introduction Previous Literature Methods and Data Results Discussions Conclusions and References A. Validation of Assumptions B. Other Models C. Pre-processing steps 4 Results The results from the model can be seen in table 1. The coefficient for time is 0.018 (SE=0.001). This implies growth in Barrier-to-Exit of 1.8% per year. This is highly significant (T=29.95, p ≪ 0.0001). The coefficient for activity level is 0.614 (SE=0.001), which is also highly significant (T=450.11, p ≪ 0.0001). A visual representation of these models can be seen in figure 4. The partial effects plot (Fox, 2003) in fig 4a shows an increase in Barrier-to-Exit from approximately 1.15 to 1.5. This translates into a growth of approximately 43% over the duration of the study. Fig. 4b shows the effect plot with residuals. This paper is available on arxiv under CC 4.0 license. Authors: (1) Jonathan H. Rystrøm. Authors: Authors: (1) Jonathan H. Rystrøm. Table of Links Abstract and Introduction Abstract and Introduction Previous Literature Previous Literature Methods and Data Methods and Data Results Results Discussions Discussions Conclusions and References Conclusions and References A. Validation of Assumptions A. Validation of Assumptions B. Other Models B. Other Models C. Pre-processing steps C. Pre-processing steps 4 Results The results from the model can be seen in table 1. The coefficient for time is 0.018 (SE=0.001). This implies growth in Barrier-to-Exit of 1.8% per year. This is highly significant (T=29.95, p ≪ 0.0001). The coefficient for activity level is 0.614 (SE=0.001), which is also highly significant (T=450.11, p ≪ 0.0001). A visual representation of these models can be seen in figure 4. The partial effects plot (Fox, 2003) in fig 4a shows an increase in Barrier-to-Exit from approximately 1.15 to 1.5. This translates into a growth of approximately 43% over the duration of the study. Fig. 4b shows the effect plot with residuals. This paper is available on arxiv under CC 4.0 license. This paper is available on arxiv under CC 4.0 license. available on arxiv

Quantifying Preference Manipulation: Growth Trends in Amazon's Barrier-to-Exit Over Time

Understanding Barrier-to-Exit in Amazon's Recommender System

অ্যামাজনের সুপারিশকারী সিস্টেমে বাধা-থেকে-প্রস্থান বোঝা

Verständnis der Ausstiegsbarrieren im Empfehlungssystem von Amazon

Comprender la barrera de salida en el sistema de recomendación de Amazon

Comprendre les barrières à la sortie dans le système de recommandation d'Amazon

अमेज़न की अनुशंसा प्रणाली में बैरियर-टू-एग्जिट को समझना

Amazon のレコメンデーション システムにおける退出障壁を理解する

Amazon 추천 시스템의 종료 장벽 이해

Compreendendo a barreira de saída no sistema de recomendação da Amazon

Понимание барьера выхода в рекомендательной системе Amazon

Amazon'un Tavsiye Sisteminde Çıkış Engelini Anlamak

Tìm hiểu về Rào cản thoát trong Hệ thống đề xuất của Amazon

了解亚马逊推荐系统的退出壁垒

BARRAISER

barraiser.com

-- employees

Since n.d.

COMPANY RANKING

BARRAISER

EVERGREEN INDEX #2814

Representing this company? Click here to claim and customize this page!

BarRaiser WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!

BARRAISER

barraiser.com

-- employees

Since n.d.

COMPANY RANKING

BARRAISER

EVERGREEN INDEX #2814

RELATED COMPANIESRANK

STORIES ABOUT BARRAISER

Representing this company? Click here to claim and customize this page!

BarRaiser WIKI

Apply to get your Tech Company News Page live on HackerNoon.com today!