2. Common Mistakes and Correct Mindsets for Research Attitudes

Many students may only “fulfill” the tasks from supervisors without having any motivation for conducting research. Here is the mindset of those students.

“If the supervisor tells me to do it, I will do it. I will only syntactically “fulfill” his/her instruction. If the supervisor does not tell me to do it, I will never think about it.” (see Fig. 2.1)

This kind of mindset is similar to submitting an assignment to a course lecturer or fulfilling the tasks from a boss in a company.

Bild vergrößern

However, this mindset must not be adopted for conducting research. Instead, students should regard research as something they need to be fully dedicated to. They need to love it as they love their boyfriends/girlfriends/sons/daughters. Using Fig. 2.2 as an example, even though the supervisor has provided one direction (XXX in this example), the student should think in a deeper way for how to further improve this paper, e.g., revise this paper thoroughly and continuously improve this work (add YYY in this example). Here, we need to emphasize that every student must have the following mindset.

Students should be mainly responsible for their research (not their supervisors). Supervisors only need to assist their research. In other words, students must be able to in charge of everything related to their own research, including (1) finding a research problem, (2) motivating this problem, (3) developing state-of-the-art solutions, (4) conducting experiments, (5) writing a research paper independently, and (6) dealing with comments from reviewers in top-tier venues independently. Figure 2.3 shows how to be a qualified Ph.D. student.

Bild vergrößern

Bild vergrößern

Many students may just regard conducting research as a job in a company, which means that they will only finish those tasks given by their supervisors. However, after they have finished them, they will never ask any additional questions. As an example, suppose that the supervisor has given the student the task for conducting an experiment, the student may just finish it in a few days and do nothing in the next few days (like the case in Fig. 2.4). As another example, suppose that the supervisor has asked the student to read one research paper. The student simply reads it and does not explore other related papers in the literature.

Bild vergrößern

Here, we need to emphasize that this mindset (or attitude) is wrong. Conducting research is not the same as a 9am–5pm job in a company. Research papers should be regarded as your asset. The main reasons are that (1) research papers are forever, which can be downloaded by everyone in the world, (2) your name is on those research papers, which means that everyone can know you (this is a glory for you), and (3) each top-tier paper can significantly help you find a tenure-track position/the best of the best research lab position (top-tier papers = money). Therefore, you should not simply wait for your supervisors to provide you the next task (especially when the meeting time is normally short, which does not allow a busy supervisor who has supervised many postgraduate students to provide extremely comprehensive comments). Instead, you need to lead your own research (keep pushing it) and ask for your supervisor help when you get stuck (see Fig. 2.5).

Bild vergrößern

Bild vergrößern

Many postgraduate students are normally the best of the best students during their undergraduate studies (e.g., receive the first-class honors). Therefore, some of these students may not be able to accept the fact that they cannot understand somethings in research papers (see Fig. 2.6). They may pretend to understand them during the meeting with others (most probably their supervisors). The main reason is that admitting the failure of understanding somethings may reveal their weakness, which can harm their prides. For those students, we would like to emphasize that this mindset is wrong. First, research papers in top-tier venues are written by researchers with rich experience. Therefore, it is possible that these papers are hard to be digested by a newbie, who does not equip with enough background knowledge. Second, if students pretend to understand somethings, the supervisors cannot determine the real progress of those students, which can significantly slow down the research progress.

In contrast, if those students can admit that they cannot understand the papers (see Fig. 2.7), their supervisors can easily identify where the students get stuck in order to provide corresponding suggestions for them. With these suggestions, the students can pass through the difficulties (i.e., earn more experience) and continue their research smoothly.

Bild vergrößern

Many unproductive students would frequently change from one topic to another topic when they have encountered some issues that are difficult to solve. They normally provide many “reasons” (we think these are just excuses) for changing topics (see Fig. 2.8).

Bild vergrößern

Bild vergrößern

Reason 1 (The topic is boring.): Some students may be very enthusiastic in the first few weeks after they have found their new research topics that they are eager to work on. However, after these students suffer from several difficulties in three to six months (Here are some possible examples. (1) It is hard to implement existing solutions. (2) There are plenty of related research papers that they cannot understand. (3) It is hard to develop a good solution. (4) It is hard to write a draft about this topic that can make their supervisors feel happy.), they will start feeling bored about this topic. At the same time, they can even grumble about why they choose this “stupid” topic before and start shifting from one research topic to another research topic when they see some new research papers that look fancy to them. For these students, we need to emphasize that no research topic is easy. They must encounter some difficulties when they conduct research for any topic. Even though they change to another fancy topic (i.e., avoid the painful experience of the previous topic), it is likely for them to feel bored after they encounter difficulties for the new topic. Therefore, the students should stick to this topic (especially when this topic is found by them) until they have finished working on it (by submitting a paper to a prestigious venue (see Fig. 2.9)). Otherwise, it can waste the precious time (e.g., three to six months) that they (and their supervisors) contribute before.

Reason 2 (The method is so simple.): Some students have developed a new method for their research topics and think that this method is so simple that is not worth for publication. Although their supervisors ask them to write down this method in the draft, they will not be very motivated for this task. Due to the weak motivation, they will not ask additional questions that can possibly remedy this (Here are some possible examples. (1) Can this simple method have some theoretical guarantees? (2) Can they support more useful problem settings? (Note that the method may need to be significantly revised for a new problem setting.)). For these students, we need to point out that publishing a research paper (1) does not mean we need to develop some solutions that look very complicated and are full of mathematics/pseudocodes and (2) does not mean we need to make somethings that can significantly change the world (e.g., formulate the equation \(E = mc^2\)). Furthermore, a simple method does not mean that it is not worth for publication. When you take a look for those publications in top-tier venues recently, many papers reuse ideas from those papers published in several decades ago. But why are these papers accepted? It is mainly because these papers (1) advance the state of the art in new problem settings, (2) have a good story (i.e., good abstract and introduction), (3) are well-written, and (4) are very comprehensive. In order to make these papers accepted in prestigious venues, all we need to do is to be motivated and keep asking questions to push this paper forward every day (see Fig. 2.9).

Reason 3 (The method cannot work for some datasets.): After students have run some experiments for testing their methods, they may find that the experimental results may not be good for some datasets. At that time, the students can feel very frustrated and think that their methods are not good in practice. Some students can immediately lose the motivation and keep saying somethings negative. Here are some possible examples. (1) This method does not work. (2) I need to develop a new method from scratch, which takes a lot of time, and I am not confident about this. Can I shift to the next topic? (3) I may not be suitable for conducting research. Let me find a job in industry. For these students, we need to emphasize that this is not the end of the world if the experimental results are not good. There are many approaches that can remedy this. We need to know that each method can have good performance in some datasets, while it can have weak performance in other datasets. Therefore, it is quite hard to develop a method that can achieve good performance for all datasets. Instead of designing another method from scratch, students should ask this question first. Which dataset properties are suitable for their methods? By figuring out the answers, they can select the correct datasets for conducting experiments (see Fig. 2.9). In addition, it is possible for them to develop a new theory/discover a new research direction behind that. As an example, suppose that the research question is to develop efficient algorithms for handling one computational problem. If the new method can only be efficient for handling sparse datasets compared with existing methods, the student should ask whether they can make the original (dense) dataset sparse without affecting the accuracy of solving the computational problem. He/she should not give up this method based on this kind of “weakness” (i.e., cannot efficiently handle dense datasets).

Reason 4 (The method is similar to a paper that was published several years ago.): Many students may discover that some methods in published papers are quite similar to their methods when they conduct literature review. After they find those papers that are similar to their ongoing papers, they will be very upset and think that this is the end of the world. They will also become not motivated for asking further questions and immediately give up this topic. For these students, we would like to ask them this question. Are these two papers really similar to each other? Previously, we supervise one student. In one meeting with this student, we discuss a new idea for solving a research problem. Then, the student keeps saying that this idea is the same as one previous research paper. But when we ask him which part is the same? He only says that the idea has one tiny part (which is hard for anyone to think of) that is the similar to that paper. Come on. Most of the ideas should be somehow similar to previous one. Therefore, how can this tiny part be interpreted as “the same idea”? As an example, suppose that we aim to utilize the compression approach for solving one computational problem. Then, can we say that this approach is the same as every compression solution? Obviously no. It is because we need to utilize the properties of the computational problem, which is the new thing, in order to develop our compression approach. Therefore, instead of being frustrated of the similarity between two papers and using this to attack the proposed idea, the students should try to find the differences (show the novelty) in order to attack the published paper (see Fig. 2.9).

Some students may think that conducting research is similar to some scientists in TV shows/movies, which depict that they are extremely hard to get one paper accepted. Each acceptance must be a breakthrough (e.g., from 5G to 6G, from weak AI to strong AI, or other Turing-award worthy/Nobel-prize worthy work). Therefore, they think that they need to spend three to four years to get only one paper accepted in a top-tier conference/journal. To be honest, these students may be idealistic (see Fig. 2.10). Normally, these students can feel very frustrated during the postgraduate studies especially when they see that (1) some colleagues can manage to publish three to four papers in top-tier venues per year and (2) those research papers in the literature are not really good (or even cannot reach the level that the authors claim before). Using the first author of this book as an example, he tried to reproduce the experimental results from one paper about manifold learning that was published in a very top-tier venue. At that time, he believed that every paper in a top-tier venue must be a kind of glory, which must be better than existing methods in all sides. However, after he conducted the experiments, he did not receive the good results as claimed by that paper. He was very frustrated and debugged the code for a long time but still could not obtain the results. Nine months later, he was fortunate to attend a top-tier conference with the first author of that paper. The author mentioned that there are some hidden assumptions in that paper and apologized to him. Therefore, we need to point out that a research paper (by extension to a research community) is not ideal at all. Instead of having serious complaints about the research community, we believe that those students need to reduce the expectation in their minds (i.e., accept the fact that it is not ideal).

Bild vergrößern

Many students would think that the level of a top-tier venue in computer science (e.g., SIGMOD, VLDB, ICDE, SIGKDD, and SIGIR) must be very difficult for them to reach. They may have the following types of illusion.

(1) Those reviewers are god. They must understand everything we write in our paper. Moreover, they can immediately know the level of each paper.

(2) Those reviewers from top-tier venues must be very serious for reviewing every single word of the submitted papers. They will spend several days or weeks for reviewing them and spotting all weak parts. They can feel very angry (and have bad impression of us) if those papers are very incremental. Therefore, we need to make sure that (i) the idea is novel enough and (ii) every single word must be used perfectly before we start writing a paper. Otherwise, we will be blacklisted by those venues.

(3) Only those papers which have solid mathematical foundations and proofs (just like those papers from Albert Einstein) can be accepted in these venues.

(4) Every paper must change the world. Otherwise, it is not worth for publishing in a top-tier venue.

For (1), we would like to mention that reviewers are just humans but not god. They cannot understand everything in computer science. In addition, computer science can be deemed as one of the fastest changing subjects in the world. In 2021/2022, many researchers focused on computational epidemiology/blockchains/IoT. In 2022/2023, many researchers focused on ChatGPT/LLM. Now, many researchers further conduct research on the topics of AI for science and AI for everything. Since many new concepts can be established in just a few years, it is infeasible for everyone (even for top researchers) to follow the new research. Therefore, instead of thinking that they understand majority of research papers (see Fig. 2.11), we can assume that they only understand a limited amount of research papers (see Fig. 2.12).

Bild vergrößern

Bild vergrößern

Bild vergrößern

Bild vergrößern

For (2), we would like to mention that those reviewers in top-tier venues normally hold the faculty positions from universities. As a faculty in a university, we are extremely busy and need to handle many stuffs every day. We need to (i) submit many papers to a conference (help editing papers from students), (ii) prepare teaching materials/assignments/examination papers, (iii) attend dissertation defenses (of undergraduate students/master students/Ph.D. students), (iv) write proposals, (v) apply patents, (vi) attend academic conferences/activities, (vii) write books, (viii) handle administrative issues from departments/faculties/schools, (ix) have meetings with many people (e.g., students and other professors), (x) organize conferences (e.g., being a PC chair), (xi) review papers from various conferences/journals. Furthermore, reviewers also have their own family. They (especially for female) need to take care of their children and their father/mother. Therefore, it is nonsense to think that reviewers spend several days to review only your paper so that they can spot every error/issue from it (see Fig. 2.13). In addition, they will also have no time and no point to let other people know how bad your paper is (see Fig. 2.14). As such, unless there is an unethical issue (e.g., plagiarism) in your paper, you can feel free to submit it to a top-tier venue. Of course, we do not encourage you to submit something really bad, i.e., far below the threshold (e.g., an incomplete draft), to a top-tier venue. But as long as the draft is complete and the presentation flow is good, you should give a try.

For (3), many junior students may think that only those papers with solid theoretical foundations and proofs can arouse the interest from reviewers. Therefore, when some students see that their idea is not very mathematical, they will deliberately add some complicated equations/formulas in the paper with no sense. When you ask them the reasons for adding the (unrelated) equations, they will simply say that (i) the technical part is too simple and (ii) these equations look cool. To be honest, they are still too young if they really do something like this. A research paper is to convey scientific information between different scientists. Therefore, the most important part is to tell a good story and show how your idea can advance the state of the art. It is completely fine that a paper can have no equation if it can still be understood by others. Suppose that the technical part looks simple. What you need to do is to dig into the problem and conduct further research. If you really want to add an equation, you need to know the necessity of adding it (but not to say that it is cool).

Bild vergrößern

For (4), many students may think that all papers in a top-tier venue must be the seminal papers, which will definitely change the world (see Fig. 2.15). However, this mindset is not correct. To be honest, majority of research papers in a top-tier venue are not very useful. Some of them even will not be mentioned in the future. Using the AAAI conference as an example, this conference accepts more than 1000 papers each year. Do you really think that 1000 seminal directions (like ChatGPT/Deepseek) can be established for one year? The answer is obviously no. Otherwise, we can achieve the goal of strong AI in a few years. Therefore, the ideas from most of these papers are in fact incremental, which only advance the field in a very limited way. However, the presentation of those papers must reach the level of a top-tier venue. As such, as long as the student has acquired the correct presentation skills, he/she can start writing papers (without being afraid of whether the idea is incremental). Some students may argue that these incremental papers are not worth for publications. Then, we will say that many seminal papers are built on top of those incremental papers (see Fig. 2.16). It is very hard for students to directly work on a seminal paper in the early stage since they do not acquire enough skills (especially for those junior students). Therefore, what they need to do is to work on some (probably incremental) research papers first (e.g., Paper 1 in Fig. 2.16). Based on the successful experience, they can explore more by asking more research questions. Ultimately, it is possible for students to achieve the impactful work. However, some students may still insist for working on seminal research without enough skills in the very early stage. Then, those students are hard to learn more skills (e.g., paper presentation skills, thinking skills), cannot even get one paper accepted (or even write a complete paper), and may be, unfortunately, kicked out from research labs.

Bild vergrößern

Many students may think that they will need to find the faculty positions in other universities. Therefore, they would like to hide some ideas so that they can publish them in the new positions (in order to make their CV better for tenure promotion). However, this mindset is only from a weak researcher (see Fig. 2.17). The main reason for a weak researcher to have this mindset is that he/she thinks that the topic space is similar to a ball with boundary. Once he/she has finished one topic, the space is reduced. Therefore, he/she has the concern for whether all topics can be done so that he/she cannot have enough papers for tenure promotion in a new university.

However, if a student is well trained to be a strong (productive) researcher, he/she does not have this concern. The main reason is that he/she regards the topic space to be a tree structure with no boundary (see Fig. 2.17). As such, once he/she has written more papers, he/she can ask more questions, which correspond to different branches of the tree structure, in order to have more research topics in the future.

Bild vergrößern

Hence, instead of worrying about the future tenure promotion, we strongly suggest that those weak researchers should improve their questioning skills. As an example, they should ask more research questions when they write each research paper. As another example, they should write more papers in order to practice these questioning skills (Practice makes perfect!). Once these skills are improved, they will not have this concern because they are no longer the weak researcher anymore.

Many students who would like to attend the postgraduate studies can normally have good academic records during the undergraduate studies (e.g., obtain high GPAs or get ACM competition awards). Therefore, some students may think that they are very good so that they refuse to listen to others (e.g., supervisors, postdoctoral researchers, and senior students). Consider the first author of this book as an example. He has some experiences for mentoring/supervising students. Here, we refer two students that were mentored/supervised by him to be “Student A” and “Student B” in the following discussion (due to the privacy issues). To his understanding, the student A initially had the better ability, including motivation, writing skills, and new methodology proposal skills, compared with the student B. However, the student A was very stubborn and arrogant, who could not accept any feedback from others and always thought that he is right. Comparatively, the student B lacked motivation but was willing to listen to his advices. At that time, he helped the student A edit research papers but the student A refused to adopt his version and told him that he did not work on the area of the student A before (so that he was not eligible to edit the paper from the student A) (see Fig. 2.18). Therefore, the student A insisted to submit some unqualified (and unpolished) research papers to top-tier conferences, leading to rejection ultimately. On the other hand, he also edited the paper from the student B. Although he was not very happy regarding the motivation from the student B, this student listened carefully with his advices and allowed him to edit the paper (see Fig. 2.18). In the end, the student A did not get any paper accepted and the student B got one paper accepted in the top-tier venue. Moreover, during the mentoring/supervising process, he observed that the ability (knowledge) of the student B significantly increased while the ability of the student A remained the same (like Fig. 2.19).

Bild vergrößern

Based on the above discussion, we emphasize that students (especially for junior students) need to be modest and listen to others, especially for some comments of writing. The main reason is that students can be regarded as newbies in research, who do not have solid experience for writing compared with faculty members/postdoctoral researchers. Without solid achievements (e.g., publishing a first-author paper in SIGMOD, where only a few parts are edited by a mentor/supervisor), it is strongly encouraged for a student to listen to his/her mentor/supervisor.

In some research groups, the supervisors (especially for those young and energetic supervisors) may have a lot of research directions for students. Moreover, due to the tenure pressure, some supervisors can act as the “mother/father” and provide every step for their students to work on (Note that the number of successful student supervisions is an important indicator for whether a faculty member can be tenured.). Therefore, those students in the research group can simply wait for the next steps from their supervisors and can possibly have many top-tier publications (see Fig. 2.20). At that time, some students, who are from other research groups, may be envious of this. However, we would like to emphasize that this may not be a good thing since a qualified research student should not always be protected under the umbrella of his/her supervisor (see Fig. 2.20). In fact, we have seen that many students who have solid research achievements in the postgraduate studies can fail to maintain the productivity after graduation (see Fig. 2.21). The main reason is that these students still do not learn how to conduct independent research during their postgraduate studies. Therefore, the student, who wants to have the career path for being a researcher in the future, should learn how to conduct independent research (though it can be painful in the first few papers). Once they have acquired the independent research skills and graduate, they should avoid having a lot of collaborations with their supervisors and try to establish collaborations with other professors in the communities. The two main reasons are that (1) they can learn different research styles from different researchers and (2) a hiring committee of a faculty member (e.g., an assistant professor position) only recruits someone who can demonstrate the independent research skills (see Fig. 2.22). Consider the first author of this book as an example. After he obtained the Ph.D. degree from the Hong Kong Polytechnic University (supervised by (Ken) Man Lung Yiu), he became a postdoctoral researcher in the University of Hong Kong (collaborated with Reynold Cheng) and a research assistant professor in the Hong Kong Baptist University (collaborated with Yun Peng, Byron Choi, and Jianliang Xu). With many collaboration and research experiences in different universities, he learned different styles and showed independent research skills (by reducing the number of publications with his Ph.D. supervisor, i.e., Ken Yiu). Observe from Fig. 2.23 that he, who graduated in March 2019, only had a small amount of representative research papers with his supervisor from 2021 to 2025 (mainly on or before 2022).

Bild vergrößern

Bild vergrößern